Let's talk about Search Relevance

Oct 26, 2021 · București, Romania

Hello,

starting October we will launch a new type of meeting in our community: Topic based meeting. What does this mean? A subject is proposed by one of our community members and we get together (this will be limited to 20 participants meeting) and discuss/learn/share - the same member of the community taking care of the moderation and organisation of the meeting. We're just starting this in October thus I expect that in time we will fine tune the idea, but until then we propose in October to meet (physically, the location is still TBD - we are looking for an outdoor place) and discuss Search Relevance. See the details below.

October organizer and moderator: Radu Gheorghe, Software Engineer (https://www.linkedin.com/in/ragheorghe/)

Let's discuss how to approach the big&fuzzy problem of search relevance. For example, when people search for "iphone" to stop giving them iphone covers first :)

I can start by sharing some of my experience (with Elasticsearch/Solr):
- lexical matching to find the right words, tolerating typos, stemming, etc
- statistical-based relevance like TF/IDF and BM25
- combining matches from different fields (e.g. product title, description...) and boosting
- injecting external info such as sales, reviews, clicks into the relevance score
- exploration via autocomplete, did you mean, faceting

And I can also scratch the surface of:
- algorithmic boost tuning, so that we stop guessing boosts
- query classification, entity extraction to make the query more targeted
- semantic/vectorized search, to hopefully stop returning chocolate-flavored X when people search for "chocolate"
- learning to rank -> using ML to inject said sales, reviews or click info in a more flexible way

I'm sure that each E-commerce implements some of the above (or some that I've missed) in different combinations. Let's share ideas of what we tried, what worked and what we think of trying :)

Radu & Valentina

Event organizers
  • Bucharest Big Data Meetup

    This is a group for anybody interested in Big Data: what it means, what tools are there and how to use them, how others are analyzing the data and what is the value they get out of it. Meetups will be both socializing and learning ones. If you are interested about the topic join us.

    Recent Events
    More

Are you organizing Let's talk about Search Relevance ?

Claim the event and start manage its content.

I am the organizer
Social
Topics
Rating

based on 0 reviews

Featured Events