What can we do to improve findability?

Question

For those who don't know me, I'm Grace Note, a Community Manager for Stack Exchange Inc. It's a pleasure to meet you.

Every two months into a beta site's lifetime, the Community Team receives a set of questions from that period of time, which is used as part of an evaluation of the site's progress and quality. One of the major focuses of this evaluation is how our answers are compared to what can be found elsewhere on the internet, and by extension how discoverable we as a site are.

A consistent element I noticed while performing the most recent reviews for the past four months, is that discoverability of this site appears to be very low. Now, as a Japanese language site that is on an English-speaking network, this is already a difficult place to break into - you have to deal with managing two separate languages and the key elements of search tend to be in Japanese. But there are still some points that could use some improvement.

Perhaps one of the major points is information in images but not in the body. This is one example, wherein the character names and not even the series name is mentioned - a lot of important context is stored within the image. Even though some of the Japanese text is in the body text, it won't show up when searching under the character's name or the series - only if you decide to search for pirates specifically. This is pretty bad.

Another example is this question. Now, I have sympathy for the asker since the kanji is slightly obscured, but it still ends that we have a question that can't be found with the specific info. The answer contains (some) of the text, but that apparently didn't improve searchability a lot - I had a lot of difficulty being able to find this one.

It's not just images, though (and in examples like this one, there's enough data in the body and the answers to find it in both the mistaken split version and the correct combined version). This question, in numerous searches explicitly for the very text that is being asked, I've barely seen our site. In fact, our own tweet of the question showed up higher on Google in one particular search. I can't explain why - it actually mystifies me, but it's still fact. This is one of but many questions where we were just not easy to find at all, when searching for the exact question.

I posted this for two reasons - I want to bring attention to the community that there is this issue, for people to be aware. The traffic to Japanese Language and Usage is actually on a decline, and I feel that this poor searchability is part of the blame of this. The second reason is that I'd like to see what the community thinks for taking steps to address this. An obvious suggestion would be to ensure more of the context that is necessary from images is contained in text. But again, images aren't the only issue - we're still being harmed either because of actual mechanical points (does the furigana hurt us in its current implementation, for example?) or from writings and styles of the question. Think on what could be done to improve the site's findability, post your findings and discuss among your friends what can be done to help this site forge its way into the internet.

Answer 1

One other thing that may be an issue that I've noticed is sometimes there are vague titles to questions, like this one, this one, or this one.

I don't think this is a particularly common problem, but maybe if we try to be a bit more conscience of question titles to make them a bit more search-engine friendly.

Answer 2

I wonder if we have a synonym problem.

When I search for one of my questions using the phrase "what is the difference between の and こと" I get my topic as the first hit, followed by several unrelated JLU questions.

If I try the similar search "when can I use こと instead of の?" that question is nowhere in sight. Comparing the first hit, a Yahoo answers question, my topic is missing the word "instead", but "の" and "こと" are mentioned quite a bit more in Derek's answer. The cached view has "の" and "こと" show up highlighted, so they don't seem to be words that Google excludes from my search (such as "when" or "the"). Perhaps since most of my search terms were in English, it gave greater weight to the English keywords?

At any rate, there are a lot of different ways to rephrase the questions we have here, but for which we don't show up in the list of results.

---

A second issue, perhaps related to SEO, is that the tags we use to classify questions aren't words people will be searching for when they are trying to look for these questions, unlike sites like StackOverflow or Gaming. For example, if I wanted to know why some people drop the "u" sound in 〜ます verbs, I probably wouldn't use the word "phonology" in my search terms. Yet, tagging the question with "phonology" makes perfect sense to users of the site, and so it shows up as the first word of the title. Some of the other sites have also complained about this.

Answer 3

One way of making the search results more available to Google could be to use Romaji alongside key words. Although I'm personally not a big fan of Romaji, I don't think it can be denied that many people use Romaji over Hiragana/Katakana/Kanji when searching on Google (especially for Japanese grammar points, which I think is a fairly large part of this site.) For example, I think it would be just as likely, if not more, that people will search for "what is the difference between no and koto" rather than "what is the difference between の and こと" (in which case I currently get a whole bunch of results which I think aren't as good as the answer here.)

On the other hand, if you search for e.g. "what's the difference between tonikaku and ichiou", it ranked fairly highly for me, as one of the answerers wrote the two in Romaji in their answer. (I think it's likely that it would've been higher if romaji was included in the question itself though.)

I think that's one of the reasons that Japanese at About.com does so well in Google results, as it includes Romaji as well as the Japanese script, e.g. it lists "Arigatou gozaimasu" as well as ありがとうございます in this page.

I don't know whether there could be some kind of automatic conversion to Romaji or not (if Romaji was decided upon.) As a demo, I built a proxy server which uses Mecab to convert Japanese to Romaji alongside the Japanese locally and tried it on this site, and it for the most part works, but there are a notable number of exceptions.

The other potential problem is Furigana, as has been outlined. I think it's likely that writing something as 行{​い​}きます would make things difficult for indexers because of the way Japanese sentences need to be broken up and analyzed. I don't know how Google handles <ruby> tags, but maybe actually serving ruby server-side might be an option? I'm not sure.

I don't think that text in images is a major problem for this site. As far as I know, there are only a few questions which suffer from this problem. The solution may be to edit those articles and add the text underneath as a blockquote etc.

---

Anyway, I'm not saying we should go with any of these options at this point, but I hope we can have some more dialog about where to go from here. This site, in my opinion, has higher quality answers than many other sites, and I think it would be great if more people were able to find them.

Answer 4

To add my two cents, I think we may have a problem due to a lot if sites that are like grammar libraries on the internet, that if people are searching for "Japanese Grammar", they are more likely going to find these types of sites. I mean sites like renshuu.org, Tae Kim's Guide, JGram, and Tanos.co.uk. I think they're helpful tools for reference, and I'm sure they get a lot more hits, and they'll turn up in more vague searches, like "help with Japanese grammar" or what-not.

I know I personally found this site when I searched for a very specific question.

I have no idea how something like this would work, or if it would be effective, but I think we would benefit by having some aspect of the site that would draw people who are more vague in their searches. That may also be what Troyen was saying. Again, I'm not sure how this would work.