How effective is naive Bayesian filtering for filtering spam?
I heard that spammers easily bypass them by stuffing extra non-spam-related words. What programming techniques can you use with Bayesian filters to prevent that?
Paul Graham was the guy to really introduce the idea of using Bayesian spam filtering to the web at large with his original article A Plan for Spam, back in August 2002. Then, his follow-up a year or so later introduced many of the problems that swiftly arose. These are still pretty great works on the topic.
In the second article, Graham mentions using CRM114, which works on a much wider set of patterns than just space-delimited words. CRM114 is cool, but comes without much implementation help for a spam filtering system.
There's the open-source powertools for Bayesian spam filtering like Death2Spam and SpamProbe.
I find nothing works quite like filtering mail through a Gmail account. Happy hunting.
I think for defeating the kind of spam attack you mention, the important thing is not the learning method but rather what features you train on. I use Fidelis Assis's OSBF-Lua which is a very successful filter: it keeps winning contests for spam filters. It uses Bayesian learning but I think the real reason for its success is three principles:
It trains not on single words but on sparse bigrams: a pair of words separated by 0 to 4 "don't care" words. The spammers have to put their message in somewhere and the sparse bigrams are very good at sussing them out. It even finds attachement spam!
It does extra training on message headers, because these are hard for spammers to disguise. Example: a message that originates on your network and never passes through an off-network relay host is probably not spam.
If the spam filter has low confidence about its classification, it requests input from a human. (In practice it adds a header field saying "please train me on this message"; the human can ignore the request.) This means that as the spammers evolve new techniques, your filter evolves to match.
This combination of techniques is extremely effective.
Disclaimer: I have worked with Fidelis on refactoring some of the software so that it can be used for other purposes such as classifying regular mail into groups or possibly one day trying to detect spam in blog comments and other places.
You're right, naive Bayesian filters are susceptible to Bayesian poisoning.
I use Popfile to not only sort away spam but also sort my email into categories and I find it hugely effective. It uses naive Bayesian filters.
Related
if we want to evaluate a classifier of NLP application with data that are annotated with two annotators, and they are not completely agreed on the annotation, how is the procedure?
That is, if we should compare the classifier output with just the portion of data that annotators agreed on? or just one of the annotator data? or the both of them separately and then compute the average?
Taking the majority vote between annotators is common. Throwing out disagreements is also done.
Here's a blog post on the subject:
Suppose we have a bunch of annotators and we don’t have perfect agreement on items. What do we do? Well, in practice, machine learning evals tend to either (1) throw away the examples without agreement (e.g., the RTE evals, some biocreative named entity evals, etc.), or (2) go with the majority label (everything else I know of). Either way, we are throwing away a huge amount of information by reducing the label to artificial certainty. You can see this pretty easily with simulations, and Raykar et al. showed it with real data.
What's right for you depends heavily on your data and how the annotators disagree; for starters, why not use only items they agree on and see what then compare the model to the ones they didn't agree on?
What approach can I use to predict the nationality of a person from the surname?
I have a huge list of texts and surnames of authors. I would like to identify which texts have been written by latin-language speakers and which texts have been written by native english speakers, in order to study if certain writing style patterns are different in one group compared to the other.
I have looked in google and in pubmed for a database of surnames, but I could not find any accessible for free. Another approach is to use some regexs, for example ".*ez" to identify some hispanic surnames such as 'rodriguez', but it doesn't get me very far.
Do you have any suggestion? Since I will manually revise all the associations after making the prediction, I don't need a great accuracy, but any help or idea will be welcome.
I don't think you can do this with any degree of reliability. A Rodriguez may well have a Spanish origin name, but could well have been born and brought up anywhere. They could be second generation British, and never have had Spanish spoken around them, and so come into the category of Native English speaker.
If Actual authors then maybe you can spider amazon and check their 'Author information' details?
I don't think you can guess. E.g. Irish last names - there are an estimated 80,000,000 people with Irish heritage however on 4.5 million of these live in Ireland/went through Irish education.
There is no meaningful way to do this. There is no reason why people with hispanic names cannot be native english speakers.
If you are going to revise it anyway, why not use the data you have?
Assuming you are intending on doing a programmatic comparison of the texts, you have to manually categorize the texts. Incorrect guesses would likely lead you to build a broken algorithm for textual analysis. This will be especially problematic with machine learning, such as artificial neural networks.
I am working with some really large databases of newspaper articles, I have them in a MySQL database, and I can query them all.
I am now searching for ways to help me tag these articles with somewhat descriptive tags.
All these articles is accessible from a URL that looks like this:
http://web.site/CATEGORY/this-is-the-title-slug
So at least I can use the category to figure what type of content that we are working with. However, I also want to tag based on the article-text.
My initial approach was doing this:
Get all articles
Get all words, remove all punctuation, split by space, and count them by occurrence
Analyze them, and filter common non-descriptive words out like "them", "I", "this", "these", "their" etc.
When all the common words was filtered out, the only thing left is words that is tag-worthy.
But this turned out to be a rather manual task, and not a very pretty or helpful approach.
This also suffered from the problem of words or names that are split by space, for example if 1.000 articles contains the name "John Doe", and 1.000 articles contains the name of "John Hanson", I would only get the word "John" out of it, not his first name, and last name.
Automatically tagging articles is really a research problem and you can spend a lot of time re-inventing the wheel when others have already done much of the work. I'd advise using one of the existing natural language processing toolkits like NLTK.
To get started, I would suggest looking at implementing a proper Tokeniser (much better than splitting by whitespace), and then take a look at Chunking and Stemming algorithms.
You might also want to count frequencies for n-grams, i.e. a sequences of words, instead of individual words. This would take care of "words split by a space". Toolkits like NLTK have functions in-built for this.
Finally, as you iteratively improve your algorithm, you might want to train on a random subset of the database and then try how the algorithm tags the remaining set of articles to see how well it works.
You should use a metric such as tf-idf to get the tags out:
Count the frequency of each term per document. This is the term frequency, tf(t, D). The more often a term occurs in the document D, the more important it is for D.
Count, per term, the number of documents the term appears in. This is the document frequency, df(t). The higher df, the less the term discriminates among your documents and the less interesting it is.
Divide tf by the log of df: tfidf(t, D) = tf(t, D) / log(df(D) + 1).
For each document, declare the top k terms by their tf-idf score to be the tags for that document.
Various implementations of tf-idf are available; for Java and .NET, there's Lucene, for Python there's scikits.learn.
If you want to do better than this, use language models. That requires some knowledge of probability theory.
Take a look at Kea. It's an open source tool for extracting keyphrases from text documents.
Your problem has also been discussed many times at http://metaoptimize.com/qa:
http://metaoptimize.com/qa/questions/1527/what-are-some-good-toolkits-to-get-lda-like-tagging-of-my-documents
http://metaoptimize.com/qa/questions/1060/tag-analysis-for-document-recommendation
If I understand your question correctly, you'd like to group the articles into similarity classes. For example, you might assign article 1 to 'Sports', article 2 to 'Politics', and so on. Or if your classes are much finer-grained, the same articles might be assigned to 'Dallas Mavericks' and 'GOP Presidential Race'.
This falls under the general category of 'clustering' algorithms. There are many possible choices of such algorithms, but this is an active area of research (meaning it is not a solved problem, and thus none of the algorithms are likely to perform quite as well as you'd like).
I'd recommend you look at Latent Direchlet Allocation (http://en.wikipedia.org/wiki/Latent_Dirichlet_allocation) or 'LDA'. I don't have personal experience with any of the LDA implementations available, so I can't recommend a specific system (perhaps others more knowledgeable than I might be able to recommend a user-friendly implementation).
You might also consider the agglomerative clustering implementations available in LingPipe (see http://alias-i.com/lingpipe/demos/tutorial/cluster/read-me.html), although I suspect an LDA implementation might prove somewhat more reliable.
A couple questions to consider while you're looking at clustering systems:
Do you want to allow fractional class membership - e.g. consider an article discussing the economic outlook and its potential effect on the presidential race; can that document belong partly to the 'economy' cluster and partly to the 'election' cluster? Some clustering algorithms allow partial class assignment and some do not
Do you want to create a set of classes manually (i.e., list out 'economy', 'sports', ...), or do you prefer to learn the set of classes from the data? Manual class labels may require more supervision (manual intervention), but if you choose to learn from the data, the 'labels' will likely not be meaningful to a human (e.g., class 1, class 2, etc.), and even the contents of the classes may not be terribly informative. That is, the learning algorithm will find similarities and cluster documents it considers similar, but the resulting clusters may not match your idea of what a 'good' class should contain.
Your approach seems sensible and there are two ways you can improve the tagging.
Use a known list of keywords/phrases for your tagging and if the count of the instances of this word/phrase is greater than a threshold (probably based on the length of the article) then include the tag.
Use a part of speech tagging algorithm to help reduce the article into a sensible set of phrases and use a sensible method to extract tags out of this. Once you have the articles reduced using such an algorithm, you would be able to identify some good candidate words/phrases to use in your keyword/phrase list for method 1.
If the content is an image or video, please check out the following blog article:
http://scottge.net/2015/06/30/automatic-image-and-video-tagging/
There are basically two approaches to automatically extract keywords from images and videos.
Multiple Instance Learning (MIL)
Deep Neural Networks (DNN), Recurrent Neural Networks (RNN), and the variants
In the above blog article, I list the latest research papers to illustrate the solutions. Some of them even include demo site and source code.
If the content is a large text document, please check out this blog article:
Best Key Phrase Extraction APIs in the Market
http://scottge.net/2015/06/13/best-key-phrase-extraction-apis-in-the-market/
Thanks, Scott
Assuming you have pre-defined set of tags, you can use the Elasticsearch Percolator API like this answer suggests:
Elasticsearch - use a "tags" index to discover all tags in a given string
Are you talking about the name-entity recognition ? if so, Anupam Jain is right. it;s research problem with using deep learning & CRF. In 2017, the name-entity recognition problem is force on semi-surprise learning technology.
The below link is related ner of paper:
http://ai2-website.s3.amazonaws.com/publications/semi-supervised-sequence.pdf
Also, The below link is key-phase extraction on twitter:
http://jkx.fudan.edu.cn/~qzhang/paper/keyphrase.emnlp2016.pdf
I am curious what are the methods / approaches to overcome the "cold start" problem where when a new user or an item enters the system, due to lack of info about this new entity, making recommendation is a problem.
I can think of doing some prediction based recommendation (like gender, nationality and so on).
You can cold start a recommendation system.
There are two type of recommendation systems; collaborative filtering and content-based. Content based systems use meta data about the things you are recommending. The question is then what meta data is important? The second approach is collaborative filtering which doesn't care about the meta data, it just uses what people did or said about an item to make a recommendation. With collaborative filtering you don't have to worry about what terms in the meta data are important. In fact you don't need any meta data to make the recommendation. The problem with collaborative filtering is that you need data. Before you have enough data you can use content-based recommendations. You can provide recommendations that are based on both methods, and at the beginning have 100% content-based, then as you get more data start to mix in collaborative filtering based.
That is the method I have used in the past.
Another common technique is to treat the content-based portion as a simple search problem. You just put in meta data as the text or body of your document then index your documents. You can do this with Lucene & Solr without writing any code.
If you want to know how basic collaborative filtering works, check out Chapter 2 of "Programming Collective Intelligence" by Toby Segaran
Maybe there are times you just shouldn't make a recommendation? "Insufficient data" should qualify as one of those times.
I just don't see how prediction recommendations based on "gender, nationality and so on" will amount to more than stereotyping.
IIRC, places such as Amazon built up their databases for a while before rolling out recommendations. It's not the kind of thing you want to get wrong; there are lots of stories out there about inappropriate recommendations based on insufficient data.
Working on this problem myself, but this paper from microsoft on Boltzmann machines looks worthwhile: http://research.microsoft.com/pubs/81783/gunawardana09__unified_approac_build_hybrid_recom_system.pdf
This has been asked several times before (naturally, I cannot find those questions now :/, but the general conclusion was it's better to avoid such recommendations. In various parts of the worls same names belong to different sexes, and so on ...
Recommendations based on "similar users liked..." clearly must wait. You can give out coupons or other incentives to survey respondents if you are absolutely committed to doing predictions based on user similarity.
There are two other ways to cold-start a recommendation engine.
Build a model yourself.
Get your suppliers to fill in key information to a skeleton model. (Also may require $ incentives.)
Lots of potential pitfalls in all of these, which are too common sense to mention.
As you might expect, there is no free lunch here. But think about it this way: recommendation engines are not a business plan. They merely enhance the business plan.
There are three things needed to address the Cold-Start Problem:
The data must have been profiled such that you have many different features (with product data the term used for 'feature' is often 'classification facets'). If you don't properly profile data as it comes in the door, your recommendation engine will stay 'cold' as it has nothing with which to classify recommendations.
MOST IMPORTANT: You need a user-feedback loop with which users can review the recommendations the personalization engine's suggestions. For example, Yes/No button for 'Was This Suggestion Helpful?' should queue a review of participants in one training dataset (i.e. the 'Recommend' training dataset) to another training dataset (i.e. DO NOT Recommend training dataset).
The model used for (Recommend/DO NOT Recommend) suggestions should never be considered to be a one-size-fits-all recommendation. In addition to classifying the product or service to suggest to a customer, how the firm classifies each specific customer matters too. If functioning properly, one should expect that customers with different features will get different suggestions for (Recommend/DO NOT Recommend) in a given situation. That would the 'personalization' part of personalization engines.
An initial draft of requirements specification has been completed and now it is time to take stock of requirements, review the specification. Part of this process is to make sure that there are no sizeable gaps in the specification. Needless to say that the gaps lead to highly inaccurate estimates, inevitable scope creep later in the project and ultimately to a death march.
What are the good, efficient techniques for pinpointing missing and implicit requirements?
This question is about practical techiniques, not general advice, principles or guidelines.
Missing requirements is anything crucial for completeness of the product or service but not thought of or forgotten about,
Implicit requirements are something that users or customers naturally assume is going to be a standard part of the software without having to be explicitly asked for.
I am happy to re-visit accepted answer, as long as someone submits better, more comprehensive solution.
Continued, frequent, frank, and two-way communication with the customer strikes me as the main 'technique' as far as I'm concerned.
It depends.
It depends on whether you're being paid to deliver what you said you'd deliver or to deliver high quality software to the client.
If the former, simply eliminate ambiguity from the specifications and then build what you agreed to. Try to stay away from anything not measurable (like "fast", "cool", "snappy", etc...).
If the latter, what Galwegian said + time or simply cut everything not absolutely drop-dead critical and build that as quickly as you can. Production has a remarkable way of illuminating what you missed in Analysis.
evaluate the lifecycle of the elements of the model with respect to a generic/overall model such as
acquisition --> stewardship --> disposal
do you know where every entity comes from and how you're going to get it into your system?
do you know where every entity, once acquired, will reside, and for how long?
do you know what to do with each entity when it is no longer needed?
for a more fine-grained analysis of the lifecycle of the entities in the spec, make a CRUDE matrix for the major entities in the requirements; this is a matrix with the operations/applications as the rows and the entities as the columns. In each cell, put a C if the application Creates the entity, R for Reads, U for Updates, D for Deletes, or E for "Edits"; 'E' encompasses C,R,U, and D (most 'master table maintenance' apps will be Es). Then check each column for C,R,U, and D (or E); if one is missing (except E), figure out if it is needed. The rows and columns of the matrix can be rearranged (manually or using affinity analysis) to form cohesive groups of entities and applications which generally correspond to subsystems; this may assist with physical system distribution later.
It is also useful to add a "User" entity column to the CRUDE matrix and specify for each application (or feature or functional area or whatever you want to call the processing/behavioral aspects of the requirements) whether it takes Input from the user, produces Output for the user, or Interacts with the user (I use I, O, and N for this, and always make the User the first column). This helps identify where user-interfaces for data-entry and reports will be required.
the goal is to check the completeness of the specification; the techniques above are useful to check to see if the life-cycle of the entities are 'closed' with respect to the entities and applications identified
Here's how you find the missing requirements.
Break the requirements down into tiny little increments. Really small. Something that can be built in two weeks or less. You'll find a lot of gaps.
Prioritize those into what would be best to have first, what's next down to what doesn't really matter very much. You'll find that some of the gap-fillers didn't matter. You'll also find that some of the original "requirements" are merely desirable.
Debate the differences of opinion as to what's most important to the end users and why. Two users will have three opinions. You'll find that some users have no clue, and none of their "requirements" are required. You'll find that some people have no spine, and things they aren't brave enough to say out loud are "required".
Get a consensus on the top two or three only. Don't argue out every nuance. It isn't possible to envision software. It isn't possible for anyone to envision what software will be like and how they will use it. Most people's "requirements" are descriptions of how the struggle to work around the inadequate business processes they're stuck with today.
Build the highest-priority, most important part first. Give it to users.
GOTO 1 and repeat the process.
"Wait," you say, "What about the overall budget?" What about it? You can never know the overall budget. Do the following.
Look at each increment defined in step 1. Provide a price-per-increment. In priority order. That way someone can pick as much or as little as they want. There's no large, scary "Big Budgetary Estimate With A Lot Of Zeroes". It's all negotiable.
I have been using a modeling methodology called Behavior Engineering (bE) that uses the original specification text to create the resulting model when you have the model it is easier to identify missing or incomplete sections of the requirements.
I have used the methodolgy on about six projects so far ranging from less than a houndred requirements to over 1300 requirements. If you want to know more I would suggest going to www.behaviorengineering.org there some really good papers regarding the methodology.
The company I work for has created a tool to perform the modeling. The work rate to actually create the model is about 5 requirements for a novice and an expert about 13 requirements an hour. The cool thing about the methodolgy is you don't need to know really anything about the domain the specification is written for. Using just the user text such as nouns and verbs the modeller will find gaps in the model in a very short period of time.
I hope this helps
Michael Larsen
How about building a prototype?
While reading tons of literature about software requirements, I found these two interesting books:
Problem Frames: Analysing & Structuring Software Development Problems by Michael Jackson (not a singer! :-).
Practical Software Requirements: A Manual of Content and Style by Bendjamen Kovitz.
These two authors really stand out from the crowd because, in my humble opinion, they are making a really good attempt to turn development of requirements into a very systematic process - more like engineering than art or black magic. In particular, Michael Jackson's definition of what requirements really are - I think it is the cleanest and most precise that I've ever seen.
I wouldn't do a good service to these authors trying to describe their aproach in a short posting here. So I am not going to do that. But I will try to explain, why their approach seems to be extremely relevant to your question: it allows you to boil down most (not all, but most!) of you requirements development work to processing a bunch of check-lists* telling you what requirements you have to define to cover all important aspects of the entire customer's problem. In other words, this approach is supposed to minimize the risk of missing important requirements (including those that often remain implicit).
I know it may sound like magic, but it isn't. It still takes a substantial mental effort to come to those "magic" check-lists: you have to articulate the customer's problem first, then analyze it thoroughly, and finally dissect it into so-called "problem frames" (which come with those magic check-lists only when they closely match a few typical problem frames defined by authors). Like I said, this approach does not promise to make everything simple. But it definitely promises to make requirements development process as systematic as possible.
If requirements development in your current project is already quite far from the very beginning, it may not be feasible to try to apply the Problem Frames Approach at this point (although it greatly depends on how your current requirements are organized). Still, I highly recommend to read those two books - they contain a lot of wisdom that you may still be able to apply to the current project.
My last important notes about these books:
As far as I understand, Mr. Jackson is the original author of the idea of "problem frames". His book is quite academic and theoretical, but it is very, very readable and even entertaining.
Mr. Kovitz' book tries to demonstrate how Mr. Jackson ideas can be applied in real practice. It also contains tons of useful information on writing and organizing the actual requirements and requirements documents.
You can probably start from the Kovitz' book (and refer to Mr. Jackson's book only if you really need to dig deeper on the theoretical side). But I am sure that, at the end of the day, you should read both books, and you won't regret that. :-)
HTH...
I agree with Galwegian. The technique described is far more efficient than the "wait for customer to yell at us" approach.