What I am trying to do:
I am trying to take a list of terms and distinguish which domain they are coming from. For example "intestine" would be from the anatomical domain while the term "cancer" would be from the disease domain. I am getting these terms from different ontologies such as DOID and FMA (they can be found at bioportal.bioontology.org)
The problem:
I am having a hard time realizing the best way to implement this. Currently I am naively taking the terms from the ontologies DOID and FMA and taking difference of any term that is in the FMA list which we know is anatomical from the DOID list (which contains terms that may be anatomical such as colon carcinoma, colon being anatomical and carcinoma being disease).
Thoughts:
I was thinking that I can get root words, prefixes, and postfixes, for the different term domains and try and match it to the terms in the list. Another idea is to take more information from their ontology such as meta data or something and use this to distinguish between the terms.
Any ideas are welcome.
As a first run, you'll probably have the best luck with bigrams. As an initial hypothesis, diseases are usually noun phrases, and usually have a very English-specific structure where NP -> N N, like "liver cancer", which means roughly the same thing as "cancer of the liver." Doctors tend not to use the latter, while the former should be caught with bigrams quite well.
Use the two ontologies you have there as starting points to train some kind of bigram model. Like Rcynic suggested, you can count them up and derive probabilities. A Naive Bayes classifier would work nicely here. The features are the bigrams; classes are anatomy or disease. sklearn has Naive Bayes built in. The "naive" part means, in this case, that all your bigrams are independent of each other. This assumption is fundamentally false, but it works well in a lot of circumstances, so we pretend it's true.
This won't work perfectly. As it's your first pass, you should be prepared to probe the output to understand how it derived the answer it came upon and find cases that failed on. When you find trends of errors, tweak your model, and try again.
I wouldn't recommend WordNet here. It wasn't written by doctors, and since what you're doing relies on precise medical terminology, it's probably going to add bizarre meanings. Consider, from nltk.corpus.wordnet:
>>> livers = reader.synsets("liver")
>>> pprint([l.definition() for l in livers])
[u'large and complicated reddish-brown glandular organ located in the upper right portion of the abdominal cavity; secretes bile and functions in metabolism of protein and carbohydrate and fat; synthesizes substances involved in the clotting of the blood; synthesizes vitamin A; detoxifies poisonous substances and breaks down worn-out erythrocytes',
u'liver of an animal used as meat',
u'a person who has a special life style',
u'someone who lives in a place',
u'having a reddish-brown color']
Only one of these is really of interest to you. As a null hypothesis, there's an 80% chance WordNet will add noise, not knowledge.
The naive approach - what precision and recall is it getting you? If you setup a test case now, then you can track your progress as you apply more sophisticated methods.
I don't know what initial set you are dealing with - but one thing to try is to get your hands on annotated documents(maybe use mechanical turk). The documents need to be tagged as the domains you're looking for - anatomical or disease.
then count and divide will tell you how likely a word you encounter is to belong to a domain. With that the next step and be to tweak some weights.
Another approach (going in a whole other direction) is using WordNet. I don't know if it will be useful for exactly your purposes, but its a massive ontology - so it might help.
Python has bindings to use Wordnet via nltk.
from nltk.corpus import wordnet as wn
wn.synsets('cancer')
gives output = [Synset('cancer.n.01'), Synset('cancer.n.02'), Synset('cancer.n.03'), Synset('cancer.n.04'), Synset('cancer.n.05')]
http://wordnetweb.princeton.edu/perl/webwn
Let us know how it works out.
Related
Consider a search system where the user submits a query ‘query’ and retrieves products based on some ranking algorithm. Assume that these products are ordered according to their quality such that p_0, p_1, …, p_10 and so on.
I would like to generate vector embeddings that mimic this ranking algorithm. The closest product vector to a query vector should ideally be p_0, the next one should be p_1 and so on.
I have tried to building word2vec embeddings for products by feeding products that have appeared in the same search session as sentences. Then, I have calculated the weighted average of product vectors to find query vectors to make the query vector closer to the top result. Although the closest result is usually the best result for a given query, the subsequent results include some results that would never appear as a top result.
Is there a trick that the word2vec can learn the ranking algorithm or any other techniques that I can try? I have looked into multi-dimensional vector scaling with non-metric distances but it did not seem scalable to me for more than 100Ks of products.
There's no one trick – just iteratively improving your representations, & training set, & ranking methods to better meet your goals.
Word2vec-based representations can often help, but are still fairly simple & centered on individual words – whose senses may vary based on context & position in ways that a simple weighted-average-of-tokens fails to capture.
You may want to represent 'products' by more than just a string-of-word-tokens – to include other properties, as well. These could be scalar values like prices or various other kinds of ratings/properties, or extra synthetic labels, such as the result of other salient groupings (whether hand-edited or learned).
And even if just working with natural-language product descriptions – like product names, or descriptions, or reviews – there are other more-sophisticated text-representations that can be trained or used – such as sentence/document embeddings using deeper-networks than plain word2vec.
Most generically, if you have a bunch of quantitative representations of candidate results, and a query, and want to use some initial examples of "good" results to bootstrap more generalizable rules for scoring top results, you are attempting a "learning-to-rank" process:
https://en.wikipedia.org/wiki/Learning_to_rank
To suggest more specific steps would require a more specific description of inputs/outputs/goals, & what's been tried, and how what's been tried has failed.
For example, are your queries always just textual product names? In such a case, maybe plain keyword search is the central technology required – with things like word-vector-modelling just a tweak for handling some tough cases, like expanding the results, or adding more contrast to the rankings, when results are too few or to many.
Or, can you detect key gaps in the modeling related to exactly those cases where "results include some results that would [ideally] never appear as a top result"? If certain things like rare (poorly-modeled) words, or important qualities not yet captured in the model, seem to be to blame for such cases, that will guide the potential set of corrective changes.
i have a small system of about 1000 documents.
For each document I would like to show links to the X "most similar" documents.
However, the documents are not labeled in any way, so this would be some kind of unsupervised method.
It feels like fasttext would be a good candidate, but I cant wrap my head around how to do it when its not labeled data.
I can calculate the word vectors, although what I really need is a vector for the whole document.
The Paragraph Vector algorithm, known as Doc2Vec in libraries like Python gensim, can train a model that will give a single vector for a run-of-text, and so could be useful for your need. Note, though, that typical published work uses tens-of-thousands to millions of documents. (Just 1,000 would be a very small training set.)
You can also simply average all the word-vectors of a text together (perhaps in some weighted fashion) to get a simple, crude vector for the full text, that will often somewhat work for this purpose. (You could use word-vectors from classi word2vec or FastText for this purpose.)
Similarly, if you have word-vectors but not full doc-vectors, there's a technique called "Word Mover's Distance" that calculates a word-vector-adjusted "distance" between two texts. It often does well in highlighting near-paraphrases, though it's somewhat expensive to calculate (especially for longer texts).
In some cases, just converting all docs to their "bag of words" representation – a giant vector containing counts of words used – then ranking docs by how many words they share is a good enough similarity measure.
Also, full-text index/search frameworks, like SOLR or ElasticSearch, can sometimes take full documents as queries, giving nicly ranked results. (This often works by picking the example document's most significant words, and using those words as fuzzy full-text queries against the full document set.)
can word2vec be used for guessing words with just context?
having trained the model with a large data set e.g. Google news how can I use word2vec to predict a similar word with only context e.g. with input ", who dominated chess for more than 15 years, will compete against nine top players in St Louis, Missouri." The output should be Kasparov or maybe Carlsen.
I'ven seen only the similarity apis but I can't make sense how to use them for this? is this not how word2vec was intented to use?
It is not the intended use of word2vec. The word2vec algorithm internally tries to predict exact words, using surrounding words, as a roundabout way to learn useful vectors for those surrounding words.
But even so, it's not forming exact predictions during training. It's just looking at a single narrow training example – context words and target word – and performing a very simple comparison and internal nudge to make its conformance to that one example slightly better. Over time, that self-adjusts towards useful vectors – even if the predictions remain of wildly-varying quality.
Most word2vec libraries don't offer a direct interface for showing ranked predictions, given context words. The Python gensim library, for the last few versions (as of current version 2.2.0 in July 2017), has offered a predict_output_word() method that roughly shows what the model would predict, given context-words, for some training modes. See:
https://radimrehurek.com/gensim/models/word2vec.html#gensim.models.word2vec.Word2Vec.predict_output_word
However, considering your fill-in-the-blank query (also called a 'cloze deletion' in related education or machine-learning contexts):
_____, who dominated chess for more than 15 years, will compete against nine top players in St Louis, Missouri
A vanilla word2vec model is unlikely to get that right. It has little sense of the relative importance of words (except when some words are more narrowly predictive of others). It has no sense of grammar/ordering, or or of the compositional-meaning of connected-phrases (like 'dominated chess' as opposed to the separate words 'dominated' and 'chess'). Even though words describing the same sorts of things are usually near each other, it doesn't know categories to be able to determine that the blank must be a 'person' and a 'chess player', and the fuzzy-similarities of word2vec don't guarantee words-of-a-class will necessarily all be nearer-each-other than other words.
There has been a bunch of work to train word/concept vectors (aka 'dense embeddings') to be better at helping at such question-answering tasks. A random example might be "Creating Causal Embeddings for Question Answering with Minimal Supervision" but queries like [word2vec question answering] or [embeddings for question answering] will find lots more. I don't know of easy out-of-the-box libraries for doing this, with or without a core of word2vec, though.
I am working user behavior project. Based on user interaction I have got some data. There is nice sequence which smoothly increases and decreases over the time. But there are little discrepancies, which are very bad. Please refer to graph below:
You can also find data here:
2.0789 2.09604 2.11472 2.13414 2.15609 2.17776 2.2021 2.22722 2.25019 2.27304 2.29724 2.31991 2.34285 2.36569 2.38682 2.40634 2.42068 2.43947 2.45099 2.46564 2.48385 2.49747 2.49031 2.51458 2.5149 2.52632 2.54689 2.56077 2.57821 2.57877 2.59104 2.57625 2.55987 2.5694 2.56244 2.56599 2.54696 2.52479 2.50345 2.48306 2.50934 2.4512 2.43586 2.40664 2.38721 2.3816 2.36415 2.33408 2.31225 2.28801 2.26583 2.24054 2.2135 2.19678 2.16366 2.13945 2.11102 2.08389 2.05533 2.02899 2.00373 1.9752 1.94862 1.91982 1.89125 1.86307 1.83539 1.80641 1.77946 1.75333 1.72765 1.70417 1.68106 1.65971 1.64032 1.62386 1.6034 1.5829 1.56022 1.54167 1.53141 1.52329 1.51128 1.52125 1.51127 1.50753 1.51494 1.51777 1.55563 1.56948 1.57866 1.60095 1.61939 1.64399 1.67643 1.70784 1.74259 1.7815 1.81939 1.84942 1.87731
1.89895 1.91676 1.92987
I would want to smooth out this sequence. The technique should be able to eliminate numbers with characteristic of X and Y, i.e. error in mono-increasing or mono-decreasing.
If not eliminate, technique should be able to shift them so that series is not affected by errors.
What I have tried and failed:
I tried to test difference between values. In some special cases it works, but for sequence as presented in this the distance between numbers is not such that I can cut out errors
I tried applying a counter, which is some X, then only change is accepted otherwise point is mapped to previous point only. Here I have great trouble deciding on value of X, because this is based on user-interaction, I am not really controller of it. If user interaction is such that its plot would be a zigzag pattern, I am ending up with 'no user movement data detected at all' situation.
Please share the techniques that you are aware of.
PS: Data made available in this example is a particular case. There is no typical pattern in which numbers are going to occure, but we expect some range to be continuous with all the examples. Solution I am seeking is generic.
I do not know how much effort you want to involve in this problem but if you want theoretical guaranties,
topological persistence seems well adapted to your problem imho.
Basically with that method, you can filtrate local maximum/minimum by fixing a scale
and there are theoritical proofs that says that if you sampling is
close from your function, then you extracts correct number of maximums with persistence.
You can see these slides (mainly pages 7-9 to get the idea) to get an idea of the method.
Basically, if you take your points as a landscape and imagine a watershed starting from maximum height and decreasing, you have some picks.
Every pick has a time where it is born which is the time where it becomes emerged and a time where it dies which is when it merges with an higher pick. Now a persistence diagram pictures a point for every pick where its x/y coordinates are its time of birth/death (by assumption the first pick does not die and is not shown).
If a pick is a global maximal, then it will be further from the diagonal in the persistence diagram than a local maximum pick. To remove local maximums you have to remove picks close to the diagonal. There are fours local maximums in your example as you can see with the persistence diagram of your data (thanks for providing the data btw) and two global ones (the first pick is not pictured in a persistence diagram):
If you noise your data like that :
You will still get a very decent persistence diagram that will allow you to filter local maximum as you want :
Please ask if you want more details or references.
Since you can not decide on a cut off frequency, and not even on the filter you want to use, I would implement several, and let the user set the parameters.
The first thing that I thought of is running average, and you can see that there are so many things to set, to get different outputs.
I'm developing a Weka app like Akinator by using the j48 method.
Sample:
http://jbossews-vdoctor.rhcloud.com/doctor
The following is the app's table definition and sample data
qa means question id(Please refer the master which can be set by user) + answer(1:Yes, 2: I don't know, 3: No).
1 line per 1 question & answer.
id,qa,class
A,13,1
A,23,1
B,13,2
B,21,2
The point is to find a way to select the question which can maximize the entropy.
Currently this app is regarding first node id of decision tree as the best question.
And then it narrows down the options by this elimination way.
But the accuracy was too bad to run correctly so I'd like to improve it.
I noticed that the qa column was identified as numeric so it could not build the correct decision tree.
I am confused what I should do for improvement. Dataset? Table definition? Logic?
This is quite a broad question that you are asking, and without code or a clear understanding of the problem it is quite difficult to answer, but I'll give some tips for improvement:
Table Definition
What may have made more sense here is to have an attribute for each question, instead of using a single instance per question. For Example, instead of id, qa and class, you could have A, B, C, D, E, F and Disease. (I believe there were six questions, and naming each attribute would be recommended instead of A-F)
Dataset
You will need at least as many cases as there are diseases, if not more for defining multiple subsets of the problem space for the same disease. There are likely cases where some questions are irrelevant or missing, and the model may need to handle such situations.
Logic
In such a case, you might be able to do the questionnaire by starting with the root node and asking questions until you reach the estimated class. This way, you can ask from node to node until a class is reached.
I hope this helps in improving your existing model.
NOTE: I tried your questionnaire and answered No to all of your questions, and I strangely ended up with Trichomoniasis. Perhaps there could be a 'No Disease' category for your training data also.
My nominal qa data is building such a decision tree by binary split.
actually this structure won't make sense because there is tree at only one side. When qa equal 23 it would be always '3' answer. It's irrational.
http://www.fastpic.jp/viewer.php?file=2693704973.jpg
You should first reformat your features to get all possible questions A,B,C,D... as binary features and your final answer (ie. what to guess) as target class if you want your tree to get a sequence of questions reaching to your answer. Your data will certainly be sparse (many questions without data/answer).
By the way, a binary tree is not the right ML structure and algorithm to build an Akinator like or 20Q/Guess-who. Please look some suggestions here: https://stats.stackexchange.com/questions/6074/akinator-com-and-naive-bayes-classifier