The linguistic boosting factor. Type of extensions: QfufFilteredByXfOneSe (qfuf, filtered by dssm-model XfOneSe). Aggregation by top-10 (by factor value) extensions. Weighted sum of factor weights. Normalized by total weight of extensions. Weighted stream aggregation of Url, Title, Body, CorrectedCtr, LongClick, OneClick, BrowserPageRank, SplitDwellTime, SamplePeriodDayFrc, SimpleClick, YabarVisits, YabarTime. Word weights aggregation algorithm: Bm15FLog (Bm15 aggregation of word occurrence logarithms). The normalization coefficient is 0.001.
tags
annotation nofilter
doc
doc text
lingboost
text machine
annotation filter
dynamic
user
url text
user search
neural
undocumented
nn over features use
userfeat
userfeat multi period
userfeat fieldset2
formula 2245 dep 2
deprecated
Details
index
1851
cppName
FI_QFUF_FILTERED_BY_XF_ONE_SE_TOP_SUM_WF_SUM_W_FIELD_SET2_BM15FLOG_K0001
name
QfufFilteredByXfOneSeTopSumWFSumWFieldSet2Bm15FLogK0001
ticket
FACTOR-2078
description
Фактор лингвистического бустинга. Тип расширений: QfufFilteredByXfOneSe (qfuf, фильтрованные по dssm-модели XfOneSe). Агрегация по топ-10 (по значению фактора) расширениям. Взвешенная сумма весов факторов. Нормированно на суммарный вес расширений. Взвешенное объединение стримов Url, Title, Body, CorrectedCtr, LongClick, OneClick, BrowserPageRank, SplitDwellTime, SamplePeriodDayFrc, SimpleClick, YabarVisits, YabarTime. Алгоритм агрегации весов слов: Bm15FLog (Bm15 агрегация логарифмов встречаемости слов). Коэффициент нормализации 0.001.
tags
TG_ANNOTATION_NOFILTER,TG_DOC,TG_DOC_TEXT,TG_LINGBOOST,TG_TEXT_MACHINE,TG_ANNOTATION_FILTER,TG_DYNAMIC,TG_USER,TG_URL_TEXT,TG_USER_SEARCH,TG_NEURAL,TG_UNDOCUMENTED,TG_NN_OVER_FEATURES_USE,TG_USERFEAT,TG_USERFEAT_MULTI_PERIOD,TG_USERFEAT_FIELDSET2,TG_FORMULA_2245_DEP_2,TG_DEPRECATED
useArtifact
[object Object]
implementationTime
2019-03-07
authors
mihaild
responsibles
mihaild,gotmanov
The ranking factor information on this page is based on the
leaked source code repository
of the Yandex Search Engine. For the purpose of this project, the source was
transformed into JSON and descriptions were translated into English. The source
is publicly available
devidw/yandex-ranking-factors .