BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//bbuzz22//talk//YEHRTE
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-bbuzz22-YEHRTE@pretalx.com
DTSTART;TZID=CET:20220613T172000
DTEND;TZID=CET:20220613T180000
DESCRIPTION:Over the decades\, information retrieval has been dominated by 
 classical methods such as BM25. These lexical models are simple and effect
 ive yet vulnerable to vocabulary mismatch. With the introduction of pre-tr
 ained language models such as BERT and its relatives\, deep retrieval mode
 ls have achieved superior performance with their strong ability to capture
  semantic relationships. The downside is that training these deep models i
 s computationally expensive\, and suitable datasets are not always availab
 le for fine-tuning toward the target domain.\n\nWhile deep retrieval model
 s work best on domains close to what they have been trained on\, lexical m
 odels are comparatively robust across datasets and domains. This suggests 
 that lexical and deep models can complement each other\, retrieving differ
 ent sets of relevant results. But how can these results effectively be com
 bined? And can we learn something from language models to learn new indexi
 ng methods? This talk will delve into both these approaches and exemplify 
 when they work well and not so well. We will take a closer look at differe
 nt strategies to combine them to get the best of both\, even in zero-shot 
 cases where we don't have enough data to fine-tune the deep model.\n\nThe 
 Search track is presented by OpenSource Connections
DTSTAMP:20260419T235555Z
LOCATION:Kesselhaus
SUMMARY:Hybrid search > sum of its parts? - Lester Solbakken
URL:https://pretalx.com/bbuzz22/talk/YEHRTE/
END:VEVENT
END:VCALENDAR