Ask HN: Preference? Keyword Search or Semantic Search

In cases where I have a very specific error code or issue I'm looking into I prefer keyword based search, but in most situations where my search is broad or I am not exactly sure what is causing an error, semantic search gets me to an answer quicker.

Although between the two, I think I use keyword searching more an a regular basis. Even in semantic search engines, I tend to use a lot of quotes to get some keyword based results.

I prefer semantic, but it needs to be good semantic search. If it's done badly, the semantic results can sometimes be rather noisy. With keywords, the onus is on the user to know what to ask for (but you can't always count on that being the case).

Google started out w/ algolia-style and then gradually added in semantic over time, you need large datasets for good quality semantic search

Keyword is simpler to implement and should be fine, but be sure to follow up at least the head queries to make sure they're getting to the right content. Things like tracking the query through to the eventual contact page, so your support reps can inform your search index; ex people who search for X need the info on faq page Y. It's a waste of your customers' time and your reps' time when the customer gave you a good enough search term, and you didn't give them the information they could have used to fix their problem without opening a ticket.

For self-service style portals I highly prefer a mix of keyword based search and good overall navigation, on this class of sites I usually know what I'm looking for and - at least from experience - common implementations of automatically induced semantic searches just seems to add noise without actually getting me closer to what I'm looking for.

Please note that my viewpoint is largely inspired by technical queries as opposed to generalized search though. In theory I'd love a mix of both with priority on keyword results under the presumption that semantic metadata is mostly curated by humans.

Keyword search I find works good. Semantic search, if implemented at all, should be an optional feature with special syntax for describing the semantics, rather than trying to interpret arbitrary text and then it won't work.

Keyword search, if it goes poorly, just feels like I need to work harder to find the right keyword, not an actual failure of the system. Whereas when semantic search goes bad, it is infuriating and entirely the fault of the computer (especially when I've given it the precise word I need it to look for and the computer "thinks" it "knows better")

...or at least that's how the medium-to-savvy users will tend to feel about it. In making your decisions you need to think not just about what users think/say they want, but also how they'll react to the system's failure modes.

I absolutely prefer keyword search. Semantic results are nothing but infuriating, particularly when I'm searching for something specific and the engine substitutes my query for something interpreted.

I prefer a semantic-guided keyword search. The query itself is keyword based, but having semantic based suggestions can help fill in some keywords I'm missing. It's as resource intensive as a regular semantic search, bridges the knowledge gap for users who can't fully describe their query, but gives full power to those in the know. Best of both worlds.

Semantic based search systems are really great at finding the original intent behind queries. Not everybody is good at finding the right keywords to express their ideas or intent. In such a scenario, semantic meaning based querying rewriting helps a lot.

However, semantic search is a hard machine learning problem and it requires a good volume of search queries so that they can be mapped to the results returned and mine for patterns.

What's the middle ground? You can show a machine learning based "related searches" as a hint in the sidebar. That way, you can help the user construct their search phrases.

Shamless plug: if you like Algolia but would like to host the search engine yourself and don't want to manage Elasticsearch, take a look at Typesense: https://github.com/typesense/typesense

Really don't know if its right to recommend a hacker but this pro Russian hacker literally saved me when he removed all negative reports on my credit and boosted my scores from 420 to over 700+. If you really need help, contact ARTHUR EMAIL- Quickarturhack ATgmail,com WHATSAPP +380683017209 or KIK Arturquickhack This man helped me hack 2 phones belonging to a blackmailer who wanted to tear my family apart. Arthur hacked both phones remotely and wiped both phones clean! He also helped my get rid of some bad records online. YOU WOULD BE MORE THAN GLAD YOU CAME ACROSS THIS POST.