BioKDE Challenge - win cash prize for searching on BioKDE!

b
biokde
楼主 (未名空间)


We are glad to announce the BioKDE Challenge exclusively for MITBBS members! If you can find a query that another search engine performs better than
BioKDE (https://biokde.com), you will win $100 cash prize for this query.
Here are the rules for the challenge:

1. The query should include only keywords in the title or abstracts of a paper.
2. The paper needs to be indexed by BioKDE. We have all the papers in
PubMed. You can check it by searching the article by its title at BioKDE.
3. The other search engines include, but not limited to, Google Scholar, PubMed, Semantic Scholar, and Microsoft Academic.
4. When searching at other search engines, only use the simple search
function (no advanced search function) so that the comparison is fair.
5. The paper should be very relevant to the keywords used.
6. Another search engine performs better than BioKDE if (1) it ranks the paper higher than BioKDE’s rank and (2) on BioKDE there is at least one
irrelevant paper ranked higher than that paper.
7. Each person can win up to 10 prizes.
8. BioKDE preserves the right to pause or stop the challenge at any time.

This challenge aims to identify the weaknesses of BioKDE search engine using a crowdsourcing approach. We thank all the participants in advance and we
believe together we can improve biomedical literature searching to benefit
all the scientists in the biomedical research community. https://biokde.com

b
biokde

Our budget for this round of challenge is $10,000.
Please reply to this post about the query you find.
Good luck!:)
m
missbaby

这网站名字里kde不好记,虽然很简单,如果能是某个新颖的组合词就好记了。也许是
我英语不好。。。。。哈哈
b
biokde

我们一直在用这个名字,所以就没有改。或许可以有奖征集一个。:)不过需要域名也要available,也不容易。
g
gates

随便使用了一下就发现一个问题,你这个搜索不会自动匹配同源基因。不同物种的同源基因名字很多时候是不一样的,比如用老鼠的某个基因名搜索,google学术会返回各物种的结果,你这个傻傻不知道,对生物学研究来说这个缺陷挺严重的
【 在 biokde (BioKDE) 的大作中提到: 】
: We are glad to announce the BioKDE Challenge exclusively for MITBBS
members!
: If you can find a query that another search engine performs better than
: BioKDE (https://biokde.com), you will win $100 cash prize for this query. : Here are the rules for the challenge:
: 1. The query should include only keywords in the title or abstracts of a
: paper.
: 2. The paper needs to be indexed by BioKDE. We have all the papers in
: PubMed. You can check it by searching the article by its title at BioKDE. : 3. The other search engines include, but not limited to, Google Scholar,
: PubMed, Semantic Scholar, and Microsoft Academic.
: ...................

b
biokde

同义词(synonyms)应该匹配。这个我们还没有做。正在做。其他的搜索引擎包括
google scholar同义词也做得不好。我们做好以后肯定比他们都好。:)
同源词如果匹配的话,不一定是用户想要的。你能举一个具体的例子吗?谢谢!

【 在 gates (大门) 的大作中提到: 】
: 随便使用了一下就发现一个问题,你这个搜索不会自动匹配同源基因。不同物种的同源
: 基因名字很多时候是不一样的,比如用老鼠的某个基因名搜索,google学术会返回各物
: 种的结果,你这个傻傻不知道,对生物学研究来说这个缺陷挺严重的
: members!
: a
: ,

b
biokde

大家觉得奖励不够多还是没有找到queries?增加奖金会有更多的人感兴趣吗?
m
missbaby

If you can find a query that another search engine performs better than
BioKDE (https://biokde.com)

你说的performs better具体是啥?更快?其实毫秒级的速度差别人感觉不出来。
你要是能把pubmed,google scholar上可用的各种插件用上以后的功能都做上去
builtin还是飞快那就很了不起了。

可以用biodig......挖挖挖。。。
b
biokde

这个在下面的第六条里说了:
6. Another search engine performs better than BioKDE if (1) it ranks the paper higher than BioKDE’s rank and (2) on BioKDE there is at least one
irrelevant paper ranked higher than that paper.

这里我们主要比较搜索结果的相关度(search results relevancy)。当然相关度可能没有绝对的标准。非常明显的差别大家还是不会有异议的。
比如你搜索“prostate cancer immune evasion”,https://biokde.com/search/?term=prostate+cancer+immune+evasion

在BioKDE上第一篇文章是
Immune landscape of human prostate cancer: immune evasion mechanisms and
biomarkers for personalized immunotherapy.
Mayassa J Bou-Dargham, Linlin Sha, Qing-Xiang Amy Sang, Jinfeng Zhang.
BMC Cancer, 2020 Jun 20; 20(1). PMID: 32552802 Free PMC article

这个是非常相关的。这可能是最相关的一篇文章。

PubMed的第一篇文章是Immune evasion in cancer: Mechanistic basis and
therapeutic strategies
Semin Cancer Biol. 2015 Dec;35 Suppl:S185-S198.

明显不如BioKDE更相关。PubMed把之前那篇最相关文章排第九位。Google scholar排在第六位。当然这个的差别不是很大。有些query差别会很大。
这里 https://biokde.com/comparison/
有一些例子。

看第一页其他的文章也会得到相似的结论。
所以就“prostate cancer immune evasion”这个query而言,BioKDE 比 PubMed和
Google Scholar的结果都好。

【 在 missbaby (请输入昵称) 的大作中提到: 】
: If you can find a query that another search engine performs better than
: BioKDE (https://biokde.com)
: 你说的performs better具体是啥?更快?其实毫秒级的速度差别人感觉不出来。
: 你要是能把pubmed,google scholar上可用的各种插件用上以后的功能都做上去
: builtin还是飞快那就很了不起了。
: 可以用biodig......挖挖挖。。。

m
merrimac

Holly cow! Non of the results from you website is even close to what I'm
looking for.

I was looking for this paper: https://www.nature.com/articles/s41586-020-
2135-x

I searched “foxa1 li j nature 2020” in both Google scholar and Bioked. It's listed in item No. 12 in Scholar, and all other listed items are also
relevant:https://scholar.google.com/scholar?hl=en&as_sdt=1%2C22&q=foxa1+li+j+nature+
2020&btnG=

而你們的結果連一點邊都沾不上:https://biokde.com/search/?term=foxa1+li+j+nature+2020
隨機在數據庫裏面取幾篇文章都比你這靠普!
b
biokde

我们现在只能用和内容相关的关键词。其他的作者,杂志和年代这一类的关键词以后再做。:)

b
biokde

之前的帖子说过,关键词一定要在文章的title或abstract里。
如果你搜“prostate cancer asian”:https://biokde.com/search/?term=prostate+cancer+asian
那篇文章排第四位。https://pubmed.ncbi.nlm.nih.gov/?term=prostate+cancer+asian
PubMed 前十没有。https://scholar.google.com/scholar?hl=en&as_sdt=0%2C10&q=prostate+cancer+
asian&btnG=
Google scholar 前十也没有。

关键词不全在文章里也可以。但是只要是和内容相关的关键词就可以。杂志,作者,年代这些和内容不相关的关键词的搜索我们以后会提供。现在就是把一个已经可以用的版本给大家先用。我们同时收集一些feedback,继续提高。那些别的搜索引擎做了十几二十几年了。我们才几个月:)。有不尽人意的地方还请多包涵。:)我们每一两个月就会有新版本推出。

年代有时候可以用左边的filter。

【 在 merrimac (不告诉你) 的大作中提到: 】
: Holly cow! Non of the results from you website is even close to what I'm
: looking for.
: I was looking for this paper: https://www.nature.com/articles/s41586-020-
: 2135-x
: I searched “foxa1 li j nature 2020” in both Google scholar and Bioked.
It'
: s listed in item No. 12 in Scholar, and all other listed items are also
: relevant:
: https://scholar.google.com/scholar?hl=en&as_sdt=1%2C22&q=foxa1+li+j+nature+
: 2020&btnG=
: 而你們的結果連一點邊都沾不上:
: ...................

b
biokde

做搜索引擎最难的是和内容相关的关键词的搜索。其他的匹配年代,作者,和杂志都相对容易得多。所以我们现在没有花很多时间在那些相对容易的问题上。

b
biokde

This event will end in three days.