Slice: video_production

All factors on L3 are applied by the matrixet formula.

All factors on L2 are applied by the matrixet formula.

Stript priority for TR is a text priority - there are all the words of the request somewhere in the document (while they pass contextual restrictions on the request, for example, both words DB in one sentence).

The presence of an accurate phrase (request text) in the header (more precisely, in the first sentence of the document).

The sum of the scales of the words of the request. Normalization: IDFSUM * 0.1 / (IDSUM * 0.1 + 1)

LVT from VH player.

The number of words in the request. Normalization: WordCount / 10.0

The document passed Softand on the restrictions of the syntactic sorcerer. Only for documents with textual relevance. For monosyllabic requests, always 1.

Text relevance (maxfreq is the frequency of the most frequent word that makes sense of the length of the document).

The request is not in Russian.

The weight of the words of the request that is in the text.

The request contains video markers: watch, film, series, etc.

The presence of a host in a request

The prediction of the turmoil of the linear model VowPalWabit, trained at the Embling Web Reformulations.

Request from class: download, watch online, play, photo, listen, etc.

Hosting video rating. Normalization is intra -hub.

Porn request.

The ratio of the number of words of the maximum general substitution of the text and the request for the words of the request.

The ratio of the number of words of the maximum general substitution of the text and the request for the words of the request with thesaurus extensions.

The largest share of the words of the request found in the leagues of the document.

The largest share of the words of the request found in the leagues of the document. Takes into account thesaurus extensions.

The largest share of the weight of the words of the request found in the leagues of the document.

The largest share of the weight of the words of the request found in the leagues of the document. Takes into account thesaurus extensions.

The number of different link languages.

BocmDouble with a weight of 0.001 on request from Corrected Clicks to Indexann.

Serial object series (normalized in [0, 1]).

The ratio of the side of the frame width / (width + height)

Valuemax for FRC on request from Corrected Clicks to Indexann.

Valueavg by FRC on request from Corrected Clicks to Indexann.

The season of a serial object, normalized on [0, 1].

BM15atten for FRC with a weight of 0.001 on request from Corrected Clicks to Indexann.

BOCMWEIGHTEDW1 UFRC with a weight of 0.01 by request from Corrected Clicks to Indexann.

The proximity of the Cluster of Instruments based on Web Reformulations

The proximity of the Cluster of films based on Web Reformulations

The proximity of the Cluster of Games based on Web Reformulations

The proximity of the Closer Closer based on Web Reformulations

Link rank the best link.

The phrasal rank of the best link.

Request in Russian.

Request in English.

BOCMDOUble by UFRC with a weight of 0.1 on request from Corrected Clicks to Indexann.

Document 1080r and has good quality.

Assessment of relevance is obtained from a duplicate of the document. It is used in training, cheerfully 0 in Prode.

Crecking the host.

Characterizes the age of the document in the context of the last 7 days. Documents over 7 days will have a zero factor.

Does the host coincide from the request with the host of the results of issuing

The number of embeds video.

The duration of the video. It is also used in the ranking of ether.

The video is an incomplete document (part 1, chapter 2).

The video is a serial object. It is marked according to the generated list of series.

The request has markers from the joke category (jokes, gum, funny).

The probable duration of the request is not less than 72 minutes, the request is not a serial and the intent Waresfilm corrects Setupids.

The request has the name of the series and a serial marker (series, series, all seasons).

The video has a film duration from 60 to 180 minutes.

The video has a serial duration from 5 to 60 minutes.

The video has a musical duration from 150 seconds to 10 minutes.

The video has a duration of 30 to 150 seconds.

The video has a trash duration from 2 to 30 seconds.

The duration of the video is unknown.

Average view depth

Some kind of crap, not a factor

Video in English is located at foreign hosting.

The video is not in Russian, is located at foreign hosting.

Video in English.

Video in Russian.

Link rank the best heading.

The phrasal rank of the best header.

Link rank the best description.

The phrasal rank of the best description.

Link rank the best link. Takes into account thesaurus extensions.

The phrasal rank of the best link. Takes into account thesaurus extensions.

Link rank the best heading. Takes into account thesaurus extensions.

The phrasal rank of the best header. Takes into account thesaurus extensions.

Link rank the best description. Takes into account thesaurus extensions.

The phrasal rank of the best description. Takes into account thesaurus extensions.

The number of braces that were put in the index.

The number of links with hits. Takes into account thesaurus extensions.

Fraction of url's path that can be covered with trigrams from the query

Similar to Urla and request by cosine metric

The share of links with hits. Takes into account thesaurus extensions.

The flag is a document trailer.

The coincidence of the ONTOID document and the object specified in the request, and a sufficiently duration of the roller for this object.

The permissible coefficient of the Bust of the document for the coincidence of the Ontoid document and the request of the type HUM, Food, Auto, Music, etc.

Prediction of Berta, trained in video relevance

Last position where the document was the worst in a pair.

The number of views of the video on the hosting.

News video.

The video is located on a trusted hosting, authorized content (for example, legal online cinema).

The authority of the host.

Share of short views(<10% and 30s)

Share of long views(>90% or 5min)

The meaning of the formula of popularity.

Score the coincidence of the Ontoid document and the main object from the request

Is recent stream

Request about the children's (0+) Mulsery.

The coincidence of the main ONTOID document and the film-serial object of the request.

Tic Hosta.

Visual quality of content on MX-model on top of content factors.

Predict of the Freshness Formula

Logdwelltime by the VEB model DSSM, trained in reformulations. It is also used in the ranking of ether.

FRC, counted for Marker-Syn normalization and extended using xFactor collaborative filtration

Increment of the number of views in 2 weeks. Characterizes the current popularity of the video.

The video has a non -zero rating on hosting.

Win (click longer than 60 seconds) on the DSSM video model trained in reformulations.

Doc is quite good to be boosted

The importance of the attractiveness of the cabinet on pair ratings.

Is the SVOD document suitable by the date of the license and the request (the same film, series series episode).

Is the AVOD document suitable by the date of the license and to the request (that film, etc.).

Random bust of relevance for rotation

The average moment of clicks.

Svod document, the date of the license is suitable ..

The average clickability of the video, calculated taking into account the position of click.

The average clickability of the video with normalization for the confidential coefficient.

The number of Ural shows.

AVOD document, the date of the license is suitable.

CTR on request.

There are shows for a couple of query.

The value of the polynomial formula of the thast.

The document is a Catchup with Yandex Hosting.

BM25 throughout the text, with standardization for the length of the document and the number of braces.

The maximum BOCM (compliance of the positions of words in the request and sentences of the document, proximity) in the text of the Links.

The video has a marked Title.

The video has a marked description.

Video has a text from a linga base.

Request language corresponds to the language of the document

The language of the document corresponds to the domain (for example, for COM.TR - this is Tur, Eng, Ger, FRN, Aze, Ara)

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. The greatest value of the factor. It is considered according to the contents of the document. The degree of covering the words of the request in the exact form.

Linguistic boosting factor. Type of extensions: QFUF. Aggregation by TOP-10 (by the value of the factor) extensions. Nimenest, balanced meaning of the factor. Normalized for the total weight of extensions. It is considered according to the contents of the document. The algorithm for aggregation of words weights is BCLMMIXPLAIN: a linear mixture of annotation BCLM weights and balanced Positionless weights of the word, then the former meters are aggregated through BM15. Normalization coefficient 10^(-5).

The proximity of the query and the DSSM model of the DSSM model, trained in relevance by distillation of the BERT model.

Coincidence of the name of the serial object in the request and document.

Season of a serial object.

Serial object series.

The coincidence of a series of a serial object in a request and a document.

The coincidence of the season of the serial object in the request and document.

The lincat database has data on the video embezzles.

Request contains a word today

LVT for browser logs for 7 days of the last 7 videos of the author.

Question about software.

Request about music.

Request about the film, the series, etc.

The clickability of the host documents on request with the normalization for the confidential coefficient. Doppelganger-norming the request and tearing video markers is used.

The share of the latest KLTIKS according to the host documents on request. Doppelganger-norming the request and tearing video markers is used.

The ratio of the number of clicks according to the documents of the host on request for all clicks on request. Doppelganger-norming the request and tearing video markers is used.

The share of deep views of the host documents on request. Doppelganger-norming the request and tearing video markers is used.

The clickability of the host documents with normalization for a confidential coefficient for 3 words of the request.

The ratio of the number of clicks according to the host documents is one of the clicks for 3 words of the request.

DiffCTR of host documents.

Weightctr documents of the host.

The share of the last clicks according to the host documents.

BocmWeighted for the relative position of Url in a chain of transitions with a weight of 0.001 on demands from webpiscus in Indexann

On the page with the video there is porn advertising. It is marked with antispam.

On the page with the video there is yellow advertising. It is marked with antispam.

On the page with the video there are priests. It is marked with antispam.

There are clickers on the page with the video. It is marked with antispam.

Famous porn player

The factor about the proximity of the request and text of the document. It differs from BCLM in that the weights of all words are considered the same. It is also used in the ranking of ether.

The request contains markers about the quality of the video - 'good quality', 'hd', '720', etc.

DBM (BM25 with machine-like words) exclusively in numbers.

The probability of an event 'Video-flow of poor quality', the Video+query model.

The probability of an event 'Video-stream of good quality (HD), model Video+Query.

The normalized number of unique words in the titles of documents raised on request, taking into account Marker-DOPP normalization.

Hosting video Semyana.

Whether the object of request and document coincided, objects are considered any film type.

The proximity of the request to the music cluster based on VPQV3

The proximity of the request to the vpqv3 -based manicure cluster

The proximity of the request to the cluster porn based on vpqv3

The proximity of the request to the cluster of computers based on VPQV3

The prediction of the porn of the linear models of VowPalWabit trained at Embeding VPQV3-V5

Bitrate video according to the video park.

Hosting video YouTube. Used in disconnecting.

Request for the film according to the Entityfinder Rules in the Byzard.

Question about software.

Request about the text.

The proximity of the request and video tack of the document according to the CV model: Target is a deep click, the step between the frames is 2.5 seconds, a maximum of 310 seconds, the picture carcass of version 8 is applied to the frame, then LSTM is used

In the marking of the video it is worth it that it is not Family-frienly, for example, through Schema.org an ISFAMILYFRIENDLY attribute.

The share of the shows of prerolls of advertising in relation to the shows

Scalar work of the vector-enlarged amount of the Embling Lingvo-Boosting extensions of the QFUF request by the Video-Reformulations model.

Quantile 0.1 For the request factor according to the XFDTSHOW linguosting data, is calculated in the LingBoostqueryFeatures Hotemaway Rules

Simple BOCM gluing Links.

BOCM for gluing Links, calculated only on the first sentences with hits and all forms of entering are considered equivalent.

The best BOCM among all links, such as Title (analogue of TOCM), calculated as follows: only sentences with hits are considered and all forms of entry are considered equivalent.

Checking for the domain RU in the request

The request was recognized as having intent to copyright objects protected by anti-Pirate memorandum.

The new DBM only in terms of gluing links (differs from DBMNUMBERS [157] only constants and completely clogs it).

The best new DBM is only by numbers in lens.

The best new DBM only by numbers, no less than 12, in links.

Host from query (for example: [cartoons on youtube]) is matched to document host. 0 = not matched, 0.5 = no specific host requested, 1 = host matched

Web fresh detector predict

Web fresh news detector predict

Prediction of the porn of the request for expanding linvobosting (VowPalWabbit Classifier)

The age of the canonical page of the video from the date of building the base (with an accuracy of days, the maximum period is 10 years).

Share of the last clicks (relative to all clicks)

The request has the name of the film or a film marker (cinema, film, melodrama).

The request has the name of the series or a serial marker (series, series, all seasons).

Is the request a wide film? For example, the 'Best Comedies of 2013' and 'Watch Online Films about Love' are wide film queries.

Corrected FRC, click := approximated (play duration/duration)

DIFFCTR Document on request with marker-peropelgangan normalization.

The average FRC document on request with the marker-peropelgangan normalization all over the whole of generalizations.

WeightCTR document on request with marker-domelgangeric normalization.

The share of the latest clicks on the document on request with marker-peropelgangan normalization.

In the request, a serial name was highlighted.

An accurate serial request.

Queryurlfrc [144], calculated by three words with the largest weight

The share of long Dwell Time Urla.

The fate of the short dwell time is urle.

The average clickability of the host with normalization for a confidential coefficient. Agggo agency on all Urlahs of the host who appeared on the issuance.

The average clickability of the Urallov for the host, calculated taking into account the position of click. Agggo agency on all Urlahs of the host who appeared on the issuance.

The average share of the last clicks (relative to all clicks) for host. Agggo agency on all Urlahs of the host who appeared on the issuance.

The average number of shows video for the host. Agggo agency on all Urlahs of the host who appeared on the issuance.

The average depth of player views of the video for the host. Aggregaled through all the urlahs of the host, according to which there are player logs.

The average age of player views of the video for the host. Aggregaled through all the urlahs of the host, according to which there are player logs.

PCTR document on request with marker normalization; Data is used in the last 30 days.

FRC document on request with marker normalization; Data is used in the last 30 days.

The language of the document is equal to the language of the search domain.

Tiktok Player

The document is defined as Stevensov.

The share of long Dwell Time host.

The fate of short dwell time host.

Coating trigrams of Title trigrams. It is also used in the ranking of ether.

Coating trigrams of a query trigrams of Urla.

The share of coincided tags allocated in the hippo, a request in the document.

Weightctr Urla per apparition, normalized by Sinsetetam.

The factor is an analogue of LongQuery (the sum of the IDF words of the request), but with the 'correct' accounting of synonyms. Specifically, a minimum of IDF (i.e. the most frequent) of synonyms and words is selected.

Average relationship of the frequency video to Webovskaya in all words of the request

The coefficient of the variation of the pieron on the clicking of the Urls for the host, calculated taking into account the position of the click. Agggo agency on all Urlahs of the host who appeared on the issuance.

The coefficient of the variation of the pieron by the share of the last clicks (relative to all clicks) for the host. Agggo agency on all Urlahs of the host who appeared on the issuance.

Request about devices.

Request about events.

Request about people.

Request about the organization.

AnnotationMatchprediction by UFRC on request from CorrectedClicks to Indexann

Request about transport.

CTR on request about devices.

CTR on request about events.

CTR on request about films.

CTR on demands about people.

CTR on demands about music.

CTR on request about the organization.

AnnotationMatchprediction for the relative position of Url in a chain of transitions at the request from web-posts in Indexann

CTR on request about software.

CTR on request about transport.

The main components of the requesting Embling from the DSSMCTRNOMINER model. It is also used in the ranking of ether.

The main components of the requesting Embling from the DSSMCTRNOMINER model. It is also used in the ranking of ether.

The main components of the requesting Embling from the DSSMCTRNOMINER model. It is also used in the ranking of ether.

The main components of the requesting Embling from the DSSMCTRNOMINER model. It is also used in the ranking of ether.

The main components of the requesting Embling from the DSSMCTRNOMINER model. It is also used in the ranking of ether.

The main components of the requesting Embling from the DSSMCTRNOMINER model. It is also used in the ranking of ether.

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. The amount by the square of the expansion, multiplied by the value of the factor. Normalized for the total weight of extensions. It is considered according to the heading of the document. PerwordCmmaxMatchMin algorithm: At least according to the maximum of the CMMAXMATCH weight abstracts.

A balanced average duration of all clicks on request

Concerted dispersion of the duration of all clicks on request

The absolute value of the difference in the duration of the video and the average duration on request

A sign of the musicality of the document

Linguistic boosting factor. Type of extensions: QFUF. Aggregation by TOP-10 (by the value of the factor) extensions. A suspended sum of the Libra of factors. Normalized for the total weight of extensions. It is considered according to the heading of the document. The degree of covering the words of the request in the exact form.

Smart processing of coincidences/availability of the name, season and episode at the narrow -serial request

The angle in the Depth Nodes space, counted only by words (min for all)

Cloud additive

The sum of the scales of the found words found in the Title. Weights from DSSM Serpsimilarityhard.

The normalized amount of the scales of the found words found in the Title. Weights from DSSM Serpsimilarityhard.

A document that we consider porn with a filed filter

A document that we consider porn during a family filter

The absolute value of the difference in the likely duration of the video and the average duration on request

'Squads' of the query - how deeply the issuance of the issuance of this request

The maximum weight of an unenlightened word in Title. Weights from DSSM Serpsimilarityhard.

A similar factor from Web.

A player with legal content.

Player of social networks.

Pleyer with an illegal counterpart.

A similar factor from Web.

Variety of the sound path (share of various coarsened words in the Chromaprint signature)

The largest tackling of the request and the domain part of Url, normalized by the length of the request

The average, according to the value of the regWordHostClickssum factor from the web

The average, according to the value of the regwordhostyabarsum factor from the web

The dynamism of the central part of the video video

Share of rollers on a host with a film duration

Share of rollers on a host with a serial duration

Share of rollers on a host with UGC duration

Share of rollers on a host with trash duration

REGWORDHOSTCLICKSSUM Factor from Web

The number of days between the date of the document and the date specified in the request. The difference is not taken into account more than a week. Read more: https://wiki.yandex-team.ru/video/videopoisk/quality/detailedFactorsdescription

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. The sum of the scales of factors. The number of extensions. It is considered according to the contents of the document. The sum of the scales of words, balanced by the weight of the annotation, is normalized for the sum of the scales of words. Only annotations are calculated, on which the sum of words of words is more than 80%.

The number of the author’s videos normalized by the sigmoid y = x/(x+50)

Diffctr for all the author’s videos.

Weightctr for all the author’s videos.

The normalized number of views of the author’s videos.

The normalized number of the last clips for the author’s videos.

The share of long views of the author of the author in DwellTime.

The share of short views of the author of the author in Dwelltime.

The average depth of viewing the author’s videos.

The average depth of viewing the author’s videos, taking into account the pruning of the duration for long rollers.

The share of long views of the author’s videos in the player.

The share of short views of the author’s videos in the player.

The average age of views of the author’s videos.

The probability of the attitude of the document to the class animals

The probability of the attitude of the document to the entertainment class

The probability of the attitude of the document to the game class

The probability of the attitude of the document to the class of Howto

The probability of the attitude of the document to the class films

The probability of the attitude of the document to the class music

The probability of the attitude of the document to the class news

The probability of the attitude of the document to the class people

The probability of the attitude of the document to the class trailers

The dynamism of the video flow

Absolute dynamism of the video flow

The request contains a freshness marker

Scalar product of the Lingvo-Boosting vector, QFUF extensions of the Video + Query V4 model,

The proximity of the request and video flow of the document according to the CV model: Target is a deep click, the step between the frames is 2.5 seconds, a maximum of 300 seconds, the picture carcass of version 9 is used to the frame, then LSTM is used

Degree of confidence in accordance with content and metadata

The degree of porn of the host is considered based on the predictions of the classifier of porn according to documents

Requested Classifier: Request about vertical videos.

BocmDouble by clicks with a weight of 0.01 on request from Sample_period Browser Sessions to Indexann

The average depth of viewing the video.

The average depth of viewing the video, taking into account the pruning of the duration for long rollers.

The share of long views of the video in the player.

The share of short views of the video in the player.

The average age of views of the roller.

HD document

Porn host

AnnotationMatchpredical on the factor of Queryurlsemanticsimilarity, widespread through XFACTOR

QueryMatchprediction on the factor of Queryurlsemanticsimilarity, distributed via XFactor

Valuewcmavg by Queryurlsemanticsimilarity factor, widespread through XFactor

The proximity of the request and the video for the Image2Text V12 model

BocmDouble with k = 0.01 according to the factor of Queryurlsemanticsimilarity, distributed via XFactor

Linguistic boosting factor. Type of extensions: QFUF. Aggregation by TOP-10 (by the value of the factor) extensions. A suspended sum of the Libra of factors. Normalized for the total weight of extensions. It is considered according to the contents of the document. The maximum weight of the annotation, the prefix of which contains the words of the request in the same order (with an accuracy to the form).

The duration of the video is normalized by the X / (1000 + X).

There is a typo in the request

Rank 'navigation'

It launches on the basic search under the name ISHUM the maximum weight of the enclosed object of the Hum or Hum1 category in the request. (See ((http://wiki.yandex-team.ru/alekseysokirko/queryobjects SOM-OV)))). ((http://wiki.yandex-team.ru/arsengadzhikurbanov/Wares#ishum more)))))

It launches on the basic search under the name Ispicture the maximum weight of the Picture or Picture1 category of the category of the category of the category in the request. (See ((http://wiki.yandex-team.ru/alekseysokirko/queryobjects SOM-OV)))). ((http://wiki.yandex-team.ru/arsengadzhikurbanov/Wares#ispicture))))))))))))))))))

It launches on the basic search under the name ISGEO the maximum weight of the meters of the gelator in the request. A geo-object is understood as an object of the category GEO, Geo1, Geoaddr, Geoaddr1, Landmark, Landmark1 (see ((http://wiki.yandex-team.ru/alekseysokirko/queryobjects kaovsky allocation))))))))))))))))))))))))))))))). wiki.yandex-team.ru/arsengadzhikurbanov/wares Read more))

Expected number of found on request

0 or 1, depending on the presence in the request of the clearly expressed intent Need_map from the variety

0 or 1 - whether the request is matured by the tickt

Popular and more or less well-known porno-hosting

The average normalized viewing time. x/(x + 300)

Pirate player

AnnotationMatchprediction by the BCLM factor from Randomlog, distributed through XFACTOR

Bocmweightedmaxk1 on the factor of BCLM from Randomlog, distributed via XFACTOR

BM15V4W1K3 according to the factor of BCLM from Randomlog, distributed via XFACTOR

BocmWeightedk7 by the Randomlog factor distributed via XFactor

Normalized number of deleted videos on the host

The average normalized energy of the chink (according to the melsprogram)

5-Percentile Flints (per Melspektrogram)

Average 90-protocerois

The first value of the volume histogram

The ratio of the first two values ​​of the volume histogram

Median volume

The probability of recognition by a neuroclassifier by an audio cargo of English

The probability of recognition by a neuroclassifier by an audio cargo of the language is not from the list: Eng, Ru, Tur

The probability of recognition by a neuroclassifier by audio cargo of the Russian language

The probability of recognition by a neuroclassifier by audio cargo of the Turkish language

The probability of recognition by a neuroclassifier by audio roads of the Animals category

The probability of recognition by a neuroclassifier by audio roads of the Autos category

The probability of recognition by a neuroclassifier by audio roads of the Comedy category

The probability of recognition by a neuroclassifier by audio roads of the Education category

The probability of recognition by a neuroclassifier by audio roads of the HOWTO category

The probability of recognition by a neuroclassifier by audio roads of the Movies category

The probability of recognition by a neuroclassifier by audio roads of the Music category

The probability of recognition by a neuroclassifier by audio roads of the News category

The probability of recognition by a neuroclassifier by audio roads of the Sports category

The probability of recognition by a neuroclassifier by audio roads of the Tech category

The probability of recognition by a neuroclassifier by audio roads of the Travel category

The video has a trailer duration

Porn query classification result from Wizard (iad_vw flag, based on Vowpal Wabbit)

Hippo factor CommercialCategories-> Cinema

The language of the document strictly corresponds to the domain (fine English)

CTR Video-Roader, taking into account real shows

The value of the classifier of narrow video checks

The value of the classifier of wide video reviews

The value of the classifier of ultrashik video reviews

A linearly normalized number of Ural shows on request with marker normalization.

Video is from kinopoisk

The proximity of the request and video flow of the document according to the CV model: carcass V7 for each of 60 frames with a step of 5 seconds, on top of them LSTM, target - deep viewing.

Is Quick Chardu

The average value of the webtrdocquorum factor is average.

The average value of the AnncorrClicksuFrcMweightedW1 AnncorrclicksuFrclicksufris.

The average value of the Duration factor is average.

The average value of the Besttitlelr factor is average.

The average value of the Besttitlephraseweight factor is average.

The average value of the Bestlrsyn factor is average.

The average value of the LINKS factor is average.

The average value of the Linkswithhitsshare factor is average.

The average value of the show factor is average.

The average value of the BestlinkBocm factor is average.

The average value of the HostdIFFCTR factor is average.

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. Type of aggregation for extensions: the sum of the weights of factors; It is normalized for the number of extensions. FRC on request from Corrected Clicks to Indexann. PerwordCmmaxMatchMin algorithm: The minimum according to CMMMAXMATCH weight.

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. Type of aggregation for extensions: suspended amount of factors weights; normalized for the total weight of extensions. FRC on request from Corrected Clicks to Indexann. The average weight of the anntations among those in which the request was an accurate tuning.

The average value of the Bocmfull factor is average.

The average value of the Bestfirsthitsentencetocm factor is average.

The average value of the Docageindays factor is average.

The average value of the showcounthostmeanreg factor is average.

Prediction of relevance using DSSM, trained to predict Bert for verticals

The average value of the DurationDifference factor is average.

The average value of the Additionrank factor is average.

The average value of the ContentVideostic0to4AVGSPEED factor is average.

The average value of the HostserialDurationDive factor is average.

The average value of the Authorshodsnormal factor is average.

The average value of the Authorlongdwelltime factor is average.

The average value of the ContentVideoSpeed ​​factor is average.

The average value of the ContentVideoAbssSpeed ​​factor is average.

The average value of the AnnbClMbClmbocmweightedMaxk1 factor is average.

The average value of the AnnbClmbM15V4W1K3 factor is average.

The number of urls in the group for request in PRS, normalized for the number of Urlov for this request in PRS.

Porn request from manual list

Is there a translated audio road from Russian for the document

Is there a translated text from Russian for the document

Whether the document is subject to translation

AnnotationMatchprediction on Indexann Strame Vuserdata2 MaxClickSPERCENT FACTOR

Querymatchpredict on Indexann Strime Vuserdata2 MaxClickSPERCENT FACTOR

VALUEWCMAVG by Indexann Strame Vuserdata2 MaxClicksPerCent Factor

AnnotationMatchprediction on Indexann Strame Vuserdata2 Factor PCTRNEW

AnnotationMatchpredcyctionWeighted on Indexann Strame Vuserdata2 Factor PCTRNEW

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. Type of aggregation for extensions: the greatest balanced value of the factor; It is normalized for the maximum weight of expansion. The result of the recognition of the text by audio document. The algorithm for aggregation of words weights is BCLMMIXPLAIN: a linear mixture of annotation BCLM weights and balanced Positionless weights of the word, then the former meters are aggregated through BM15. Normalization coefficient 10^(-5).

WEBFRESHDATEFACTORS-> DATEISPRESENTED Hippo factor

SynonyMatchpredical on Indexann Stream Web Corrected_ctr_xfactor. Regional only for Turkey

AnnotationMatchpredicTionWeighted on Indexann Stream Web Corrected_CTR_XFACTOR. Regional only for Turkey

Querymatchprediction by Indexann Stream Web Corrected_ctr_xfactor. Regional only for Turkey

AnnotationMatchpredictionWeighted by Indexann Stream Veba One_Click_frc_xf_sp

Querymatchprediction by Indexann Stream Web One_Click_FRC_XF_SP

BM15stRICTK2 by Indexann Strip Welb One_Click_FRC_XF_SP

BCLMPLAINK3 by Indexann Strip Weln_Click_FRC_XF_SP

VALUEWCMPREDICATION ON INDEXANN STRIM VEBE LONG_CLICK_SP

BOCMDOUBLEK1 PO INDEXANN Strimu Veba Long_Click_sp

BOCMDOUBLEK5 PO INDEXANN Strimu Veba Long_Click_SP

VALUEWCMAVG by Indexann Strip Veb Simple_Click

BM15stRICTK2 by Indexann Strip Veb Simple_Click

BOCMDOUBLEK5 PO INDEXANN Strimu Veba Simple_Click

AnnotationMatchprediction on Indexann Strip Vib Bqpr_sample

VALUEWCMAVG by Indexann Strip Veb BQPR_SAmple

BCMDOBLEK1 by Indexann Strip Veb BQPR_Sample

The average depth of viewing the video, taking into account the pruning of the duration for long rollers in 5 days.

The average depth of viewing the video, taking into account the pruning of the duration for long rollers in 3 days.

The average depth of viewing the video, taking into account the pruning of the duration for long rollers in 1 day.

The share of short Dwell Time Urla in 5 days.

The share of short Dwell Time Urla in 3 days.

The average position of Urla Urla in 5 days.

Corrected Clicks in 5 days.

Corrected Clicks for 3 days.

Corrected clicks за 1 день.

The average clickability of the video, calculated taking into account the position of click in 5 days.

The average clickability of the video, calculated taking into account the position of click in 3 days.

The average normalized viewing time. x/(x + 300) in 3 days

The average clickability of the video with normalization for the confidential coefficient in 3 days.

The average depth of viewing the host, taking into account the pruning of the duration for long rollers in 5 days.

The average depth of viewing the host, taking into account the pruning of the duration for long rollers in 3 days.

The average depth of viewing the host, taking into account the pruning of the duration for long rollers in 1 day.

The share of the short Dwell Time host in 5 days.

The share of the short Dwell Time host in 3 days.

The share of short Dwell Time host in 1 day.

The average depth of viewing the author, taking into account the pruning of the duration for long rollers in 5 days.

The average depth of viewing the author, taking into account the pruning of the duration for long rollers in 3 days.

The average depth of viewing the author, taking into account the pruning of the duration for long rollers in 1 day.

The share of the short Dwell Time author in 5 days.

The share of the short Dwell Time of the author in 3 days.

The fate of the short dwell time of the author in 1 day.

The share of long Dwell Time query taking into account MarkerDopPP-Normalization.

The flag of the presence of a player for a document for this platform/user device.

The average value of the ISUGCDURATION factor in PRS

The average value of the Queryurlsemanticsimilar factor in PRS

The average value of the Bocmfull factor in PRS

The average value of the Bestfirsthitsentencetocm factor in PRS

The average value of the CorrectedmaxClicksperteMaxPercentDopp factor in PRS

The average value of the Markdoppurldiff factor in PRS

The average value of the Markdoppurlfrcmean factor in PRS

The average value of the Additionrank factor in PRS

The average value of the CATEGMUSICPROBABABIBILIT factor in PRS

The share of the shows of prerolls of advertising in relation to the shows exceeds the threshold (0.1)

The average quadrary deviation of the AnncorrClicksBocmDouble factor in PRS

The average quadracial deviation of the AnncorrClicksfrcvaluemax factor in PRS

The average quadratic deviation of the AnncorrClicksuFrcBocmDouble factor in PRS

The mid -sequential deviation of the Duration factor in PRS

The mid -sequential deviation of the Bocmfull factor in PRS

The mid -sequential deviation of the Titlerigramsinquery factor in PRS

The average deviation of the Additionrank factor in PRS

The average quadratic deviation of the CategGamesProbaboly factor in PRS

The mid -sequential deviation of the CategmoviesProbaboly factor in PRS

The mid -sequential deviation of the CategmusicProbaboly factor in PRS

The mid -sequential deviation of the CategnewSProbaboly factor in PRS

The average deviation of the PlayershortViewsNew factor in PRS

The average quadratic deviation of the Annqusmvaluewcmavg factor in PRS

The duration of the object from the ontobase, which was struck with a request

CoEF-T of Pieron Variation for the Duration factor in PRS

The weight of the query match with the object from the ontobase

CoEF-T of Pieron Variation for the Diffctr factor in PRS

CoEF-T of Pieron Variation for QueryurlsemanticSimilarity factor in PRS

COF-T Variations Pyrson Dul factor Titlecovering in PRs

COF-T variations Pyrson Dul factor CorrectedmaxClickspercentdopp in PRs

CoEF-T of Pieron Variation for the Titlerigramsinquery factor in PRS

COF-T Variations Pyrson Dul factor ismusic in PRs

CoEF-T of Pieron Variation for Additionrank factor in PRS

CoEF-T of Pieron Variation for the Categanimalsprobaboly factor in PRS

CoEF-T of Pieron Variation for the CategHowtoprobaboly factor in PRS

CoEF-T of Pieron Variation for the AnnqusmQuryMatchprediction factor in PRS

CoEF-T of Pieron Variation for PlayERviewtime factor in PRS

CoEF-T of Pieron Variation for the AnnbClmannotationMatchprediction factor in PRS

The duration of the object from the ontobase, which was struck with a request, corresponds to the duration of the video (ratio from 0.8 to 1.2)

Resid for Webtrdocquorum Factor in PRS

Resid for the Duration factor in PRS

Resid for the IsserialDurament factor in PRS

Resid for the ISUGCDURATION Factor in PRS

Resid for Titlecovering Factor in PRS

Resid for Contentvideostatic0To4AVGSPEED FACTOR

FRACT for HostshortdwellTime factor in PRS

FRACT for the Titlerigramsinquery factor in PRS

FRACT for the MarkSynurlweightctr factor in PRS

FRACT for ContentVideostic0to4AvgSpeed ​​factor in PRS

Fract Dul factor ContentVideospeed in PRS

The number of urls on the host for request in PRS, normalized for the number of Urlov for this request in PRS.

The duration of the object from the ontobase, which was struck on request, to the duration of the video

The value of porn by the classifier

Serial request contains season number

The value of the classifier Pagequaly

Bocm for factorann that does not use sent/data files

Bm15Weighted for factorann that does not use sent/data files

Bm15Plain for factorann that does not use sent/data files

BclmLight for factorann that does not use sent/data files

Time of moderation pass for videotop in given region: 0 - 7 days ago, 1 - right now

Toloka like mark from new pool

The proximity of the request and document on the Neuronet model trained on DSSM and static/request factors.

The flag is Lee Host Yandex.Chosting.

Classifier response to Video+Query V3 'Visual quality == Good'

Classifier response to Video+Query V3 'Visual quality == Norm'

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. Type of aggregation for extensions: the sum of the weights of factors; It is normalized for the number of extensions. Stream VideouunitedFactorann. Algorithm Allwcmweightedpredical.

Linguistic boosting factor. Type of extensions: QFUF. Aggregation on all extensions. Type of aggregation for extensions: suspended amount of factors weights; normalized for the total weight of extensions. Stream VideouunitedFactorann. CMMATCH80AVGVALUE algorithm.

Diffctrvideotop3d user factor by video recovery for 3 days

Weightctrvideotop3d user factor by video recovery in 3 days

User factor AverageshowpositionVideotop3d on a video recovery for 3 days

SHORTDWELTIMEVIDOTOP3D user factor by video recovery in 3 days

User factor PlayERViewtruncdepthNewVideotop3d on a video recovery for 3 days

User factor of Playarsevtimevideotop3d on a video recovery for 3 days

CorrectedCtrvideotop3d user factor by video recovery in 3 days

Diffctrvideotop1d user factor by video recovery in 1 days

Weightctrvideotop1d user factor by video recovery in 1 days

User factor AverageshowpositionVideotop1d on a video recovery in 1 day

SHORTDDWELTIMEVIDOTOP1D user factor by video recovery in 1 day

User factor PlayERViewtruncDepthNewVideotop1d for a video recovery in 1 day

User factor of PlayERViewtimevideotop1d on a video recovery in 1 day

CorrectedCtrvideotop1d user factor by video recovery in 1 day

Fullmatchprediction on Stream Playersewtimemax from browser logs

SynonyMatchpredication for Playersewtimemax Stream from browser logs

AnnotationMatchpredcyctionWeighted on a restraint of playerViewtimemax from browser logo

QueryMatchprediction on Stream Playersewtimemax from browser logs

Valuewcmavg on the stream Playervtimemax from browser logs

BCLMBLAINK3 Strim Playervtimemax from browser logs

The number of views of the video on the hosting (stat. Remap).

Ligvobusting factor. Type of extension - XFDTSHOW, stream - text, factor - BM11ALLW0, maximum value of the factor for extensions.

Ligvobusting factor. Type of extension - XFDTSHOW, stream - text, factor - BOCM11, maximum value of the factor for extensions.

Ligvobusting factor. Type of extension - XFDTSHOW, stream - description of the video, factor - pairminProximity, maximum value of the factor in extensions.

Ligvobusting factor. Type of extension - XFDTSHOW, stream is the title of the roller, the factor - BM15, the maximum balanced value of the factor for extensions.

Ligvobusting factor. Type of extension - XFDTSHOW, stream - text, factor - BM11ALLW0, maximum balanced value of the factor for extensions.

Ligvobusting factor. Type of extension - XFDTSHOW, stream - description of the video, factor - FullmatchCoveraGestartingPercent, maximum balanced value of the factor for extensions, rationing for max. The meaning of the factor.

Ligvobusting factor. Type of extension - XFDTSHOW, stream - description of the video, factor - FullmatchCoveraGestartingPercent, the maximum balanced value of the factor for extensions, standardization for the amount of weights.

Ligvobusting factor. Type of extension - XFDTSHOW, stream - description of the video, minimum weight by extensions.

Ligvobusting factor. Type of extension - XFDTSHOW, Stream - Text, Factor - FullmatchCoveRagestartingPercent, the minimum value of the factor for extensions.

Ligvobusting factor. Type of extension - XFDTSHOW, stream is the title of the video, the factor - BM11, the average balanced values ​​of the factor multiplied by weight.

Ligvobusting factor. Type of extension - xfdtshow, a bag of words, stream - a title of a video, factor - AnnotationMatchavgvalue.

Ligvobusting factor. Type of extension - xfdtshow, bag of words, stream - text, factor - cosinemaxmatch.

Ligvobusting factor. Type of extension - xfdtshow, bag of words, stream - text, factor - cosinematchweightedValue.

Increment of the number of views in 3 days. Characterizes the current popularity of the video.

Logdwelltime by the VEB model DSSM. It is also used in the ranking of ether.

HTML5 Flag at the host of the document.

Time of moderation pass for videotop in given region: 0 - 30 days ago, 1 - right now

The flag that the request came from the desktop.

The presence of clickbeite markers in the video title

Corrected CTR for issuing mobile devices.

Diffctr for mobile devices.

The average age of the video on mobile devices.

The sum of the depths of viewing video on mobile devices at the time of viewing.

Whether this document is good in the opinion of Tolockers (0.5 - lack of evaluation).

If a word from a request is found in Ann adds a normalized weight of the word to this factor 1 time.

The number of divisions by the document in FB, which returns Graph API

Number of likes from hosting

Number of likes from hosting

Normalized number of Urla duplicates on all hosts

The number of straight lines from browser logs in 8 hours

The number of visits from browser logs in 8 hours

The number of deep views from the video disast of 8 hours

The number of short views from the video file in 8 hours

The number of straight lines from browser logs in 72 hours

The number of visits from browser logs for 72 hours

The number of deep views from the video file for 72 hours

The number of short views from the video file for 72 hours

Tolocieran assessment of video intent strength at the request

Tolocieran assessment of the integration force of freshness at the request

The presence of a tolookier assessment of the integration force at the request

Logdwelltime by the VEB model DSSM.

The similarity of i2T vehicles of the cabinet and request trained on these videos

Probability of the REL label from the toloka

Probability of the NOT_REL label from the toloka

Probability of the 404 label from the toloka

Probability of host nastiness

Is online stream

Is big popular host

The similarity of the T2q vehicles of the Title and the request a la Klakhman, trained by clicks

The freshness of the serial document on attributes (> = 0.8 if the series has not yet released).

Whether the request is a large -series.

DSSM with PlayerDepth Target on the deep click pool video. It is also used in the ranking of ether.

Not a suitable series (another series, season or series).

There is a Moonwalk player

DSSM with PlayerDepth Target on the deep click of a video at L2 Stages.

The number of terms by which a document was made up by the number of terms by which the document was fucked up, plus one

The number of terms by which a document divided into the total number of terms was fucked up

Panther relevant, divided into panther relevant plus 256

Panther relevantity divided by the number of terms in the request

Panther relevantity divided by the number of words in the request

Panther relevant, divided by the number of terms by which the document was fucked

Panther relevant, divided by the number of words from the request, according to which the document was fucked

The number of words from the request for which the document

The number of words from the request for which a document divided into the total number of words

The amount of IDF words by which a document divided into a total IDF was fucked up

Position in the top panther

The value of the classifier of relevance of video queries

Frame width / 1920

Frame height / 1080

The nonstop player is connected by the document for this platform.

The probability that the VK.com host is popular for this request in accordance with the corresponding DSSM model.

The probability that the Onliner.by host is popular for this request according to the corresponding DSSM model.

The probability that the Rambler.ru host is popular for this request in accordance with the corresponding DSSM model.

The probability that the Sunhome.ru host is popular for this request according to the corresponding DSSM model.

Summary Skorov words of a request for a language model 3GRAMS-YANDEX-DIRECT.

Summary of the Skorov words of the request by the Web-Mt language model.

Linguistic boosting factor. The average weight of extensions type Querytotext.

Linguistic boosting factor. The average weight of the QFUF type extensions.

Linguistic boosting factor. Type of extensions: QFUF. Transferred the total weight of the extensions.

ADDTIME average value for the year. It is calculated in offline.

The average value of TXTHIRELSY for the year. It is calculated in offline.

The average TEXTLIKE value is for the year. It is calculated in offline.

The average value is Yabarhostavgtime2 for the year. It is calculated in offline.

The average value of Yabarurlvisitors for the year. It is calculated in offline.

The average value of QueryDowneronlyClickRate for the year. It is calculated in offline.

The average value of Dateraage for the year. It is calculated in offline.

The average value of the LonGestText for the year. It is calculated in offline.

The average value is DifferentinTernallinks for the year. It is calculated in offline.

The average value of QueryDowneronlyClickRate_Rreg is for a year. It is calculated in offline.

The average ISHUB value is for the year. It is calculated in offline.

The average ISTARGETBUSSINESSCARD for the year. It is calculated in offline.

The average value is BM25_0 on request per year. It is calculated in offline.

The average value of BOCM for the year. It is calculated in offline.

The average ISindexpage is for the year. It is calculated in offline.

The average value of queriesavgcm2 for the year. It is calculated in offline.

The average value of BrowserhostDownloadProbabolyity for the year. It is calculated in offline.

The average value of Regbrowseruserhub for the year. It is calculated in offline.

The average value is QueryTodocallsumfcountTextbM11Norm16384 for the year. It is calculated in offline.

Maintened by clicks ISMainPage value for the year. It is calculated in offline.

Main -heated clicks of the Yabarurlavgtime value for the year. It is calculated in offline.

Malpanized Dwelltime-AMI Value of Urldomainfraction for the year. It is calculated in offline.

HardPornoclassifiertype: 0 - The result is not calculated, 1 - non -hard porn,> 1 - HardpornoclassifierProbability is no less than the threshold for a given recall (2 - 0.99, 3 - 0.98, 4 - 0.95, 5 - 0.90, 6 - 0.83, 7 - 0.74)

The document contains tin (blood, open wounds, etc.)

Median dwelltai request in history. Dwelltaym is cut to 6000. The request is normalized by doppelgangers

The number of shows of the request with more than one click in history. The request is normalized by doppelgangers

The share of shows with more than one click from all shows in history. The request is normalized by doppelgangers

25% time quantile from the previous request to the current one. The request is normalized by doppelgangers

Predicted by demand and country, using a DSSM model, the length of the click from this country.

The average for the year for the year predicted using the neural network.

ADDTIME ADDTIME is predicted using a neural network for a year.

The average Txthirelesy value predicted using a neural network for the year.

The average Textlike is predicted using a neural network for the year.

The average HasnoallwordStrsy is predicted using a neural network for a year.

The average ISFORUM is predicted using a neural network for the year.

The average value of Yabarhostavgtime2 for the year for the year.

The average yabarurlvisitors is predicted using a neural network for the year.

The average value of QueryDowneronlyClickRate for the year for the year.

The average Dateraage value for the year for a year predicted using a neural network.

The average LonGestText is predicted using a neural network for the year.

The average DifferentinTernallinks for the year for the year.

The average value of QueryDowneronlyClickRate_Rreg is predicted using a neural network for a year.

The request has a voice acting studio that coincides with the voice acting studio in the document

The average ISTARGETBUSSINESSCARD is predicted using a neural network for the year.

The average BOCM value predicted using a neural network for the year.

The average ISindEXPAGE is predicted using a neural network for the year.

The average value of QueriesavGCM2 for the year for the year predicted using a neural network.

The average BrowserhostdowLoadProbabolyti is predicted using a neural network for the year.

The average value of Regbrowseruserhub for the year for a year predicted using a neural network.

The average value of QueryurlCorrectrxFactor for the year for the year.

The average value of QueryTodoCallsumfcountTextbM11Norm16384 for the year for the year.

The average value of the XFDTSHOWALSUMWFSUMWBODYMINWINDOWSIZE for the year for the year.

The value of the ISMAINPAGE with clicks predicted using the neural network with clicks on request for the year.

A mid Yabarurlavgtime value predicted using a neural network with clicks for a year.

DiffferentinTernallinks, which is predicted using a neural network, is a weighted net with clicks for a year.

The Malue Network DwellTime-AMI predicted using the neural network is the value of Urldomainfraction for the year.

LVT for browser logs in 180 days.

LVT for browser logs in 30 days.

LVT for browser logs in 7 days.

LVT for browser logs in 1 day.

Users who entered this URL on browser logs in 180 days.

Users who entered this URL on browser logs in 30 days.

Users who entered this URL on browser logs in 7 days.

Users who entered this URL on browser logs in 1 day.

LVT for browser logs in 180 days.

LVT for browser logs in 30 days.

LVT for browser logs in 7 days.

LVT for browser logs in 1 day.

Users who entered this URL on browser logs in 180 days.

Users who entered this URL on browser logs in 30 days.

Users who entered this URL on browser logs in 7 days.

Users who entered this URL on browser logs in 1 day.

Characterizes the request for the degree of change from the addition of a fixed word (number of some year), DSSM model DSSMBOOSTINGXFONESEAMSARD is used

It characterizes the request for the degree of change from the addition of a fixed word ('online' for Kirilitsa), the DSSM model DSSMBOOSTINGXFONESEAMSHARD is used. It is also used in the ranking of ether.

Characterizes the request for the degree of change from removing a fixed word ('site' for Kirilitsa), DSSM model DSSMBOOSTINGXFONESEAMSARD is used

For each word offline, the average Hasnotr meaning is calculated for 3 months. Further, in all words of the request, the maximum of this value is taken.

For each word offline, the average BCLMLITE value is calculated for 3 months. Further, in all words of the request, a minimum of this value is taken.

For each word offline, the average DBM40 value is calculated for 3 months. Further, for all non -feet, the words of the request are taken as a maximum of this value.

For each word offline, the average ISDESKTOPREQUEST value is calculated for 3 months. Further, for all non -feet, the words of the request are taken as a maximum of this value.

For each word offline, the average value of RLQAVGHASNOLLWORDSTRSYN is calculated at the request for 3 months. Further, in all words of the request, the maximum of this value is taken.

For each word offline, the average DSSMAGGRETEDANNREG value is calculated at the request for 3 months. Further, in all words of the request, the maximum of this value is taken.

For each word offline, the average meaning of MetanumurlSperhostfixed is calculated in demands in 3 months. Further, in all words of the request, the maximum of this value is taken.

For each word offline, the average value of MaxsdisnavmxqueryMax is calculated at the request for 3 months. Further, for all non -feet, the words of the request are taken as a maximum of this value.

DSSM prediction of google specificity for query

Date of issue of the document (for TV shows).

Whether the document is consistent with a wide request (wide request: = [genres] [country] [year] [age limit (0 | 6 | 12 | 16)] [category (film | series | clip)]. Example: (French film About the war for adults 2018)

Rating of KinoPoisk. The minimum rating from all object cards.

The video has a filmpois rating.

Predict formula for the relevance of a fresh document.

Random bust of relevance for rotation

In this factor (into this index), it seems to be rewritten in an unknown place, presumably from some dense Legacy Code

A sign that the document is a collection

The position of the document in L1 top. Used to define targett in teaching a dynamic L1 Top model

PCA0 from the request of the VEB -VEB model DSSM, trained in reformulations.

PCA1 from the request of the VEB -VEB model DSSM, trained in reformulations.

PCA2 from the request of the VEB model DSSM, trained in reformulations.

PCA3 from the request of the VEB -VEB model DSSM, trained in reformulations.

PCA4 from the request of the VEB -VEB model DSSM, trained in reformulations.

PCA0 from the request of the DSSM video model trained in reformulations.

PCA1 from the request of the DSSM video model trained in reformulations.

PCA2 from the request of the DSSM video model trained in reformulations.

PCA3 from the request of the DSSM video model trained in reformulations.

PCA4 from the request of the DSSM video model trained in reformulations.

PCA5 from the request of the DSSM video model trained in reformulations.

Assessor assessment age - is considered only when learning, in production is always 0, which corresponds to the current moment.

The factor for the original request. It is considered according to the contents of the document. Algorithm: AVGPERGRAMAMAVGVALueany.

The factor for the original request. It is considered according to the contents of the document. The BCLMPLANEPROXIMITY15W0SIZE1 algorithm: uses BCLM with free weighing if there are several words, if the word is one, then the sum of hits is used as a type of coincidence. Normalization coefficient 0.01.

The factor for the original request. It is considered according to the contents of the document. Algorithm for aggregation of the scales of words BOCM15. Normalization coefficient 0.01.

The factor for the original request. It is considered according to the contents of the document. The degree of covering the words of the request in the exact form.

The factor for the original request. It is considered according to the heading of the document. The degree of coating of the words of the request (all types of hits).

Hippo factor KnnrandomlogqueryFactors-> Conveorrandomlogqueryavgqfuflmaxwflinkannindicatorfullmatchvalue

Quantum geographical breadth of the city, from where a request was set

Video classifier value Video

Query-to-doc click similarity factor (query-side), with query vector initialized as bag-of-words vector

Query-to-doc click similarity factor (query-side)

Query-to-doc click similarity factor (query-side), with query vector initialized from qfuf expansions

Query-to-doc click similarity factor (query-side), with query vector initialized as bag-of-words vector, min ann weight

Max weight of a word from click similarity query vector which is not found in CS document vector

Max weight of a word from click similarity qfuf query vector which is not found in CS document vector

Query-to-doc click similarity factor (query-side), with query vector initialized as normalized idf vector

Linguistic boosting factor. Type of extensions: Originalrequest. Aggregation in all words of expansion. The sum of the square of the weight of the word (IDF), multiplied by the meaning of the factor. Normalized for the total weight of words. It is considered to be CS the vector of the document. PerwordCmmaxMatchMin algorithm: At least according to the maximum of the CMMAXMATCH weight abstracts.

Prediction of image thumb nn classifier

Is the video vertical

Prediction of the classifier of the maturity of requests for video

Prediction of relevance using DSSM, trained to predict Berta

Is license required to show a video

Did you jump

Whether fragments are highlighted inside the video

Whether the sound track is recognized as a text

Share of documents with downloaded video content on the host

Share of documents with allocated fragments inside the video on the host

The share of documents with a sound path recognized as a text on a host

Share of documents with downloaded video content for the author

Share of documents with allocated fragments inside the video for the author

The share of documents with a sound path recognized as a text for the author

Aspect of relevance for Proxima-Klean