I've seen some questions on deciding the best OCR result given output from different engines, and the answer is typically "choose the best engine".
I want, however, to capture several frames of text images, with possible temporary occlusions or temporary failures.
I'm using tesseract-ocr with python-tesseract.
Considering the OCR outputs of the last N frames, I want to decide what is the best result (line by line, for simplicity).
For example, for N=3, we could use a median filtering:
ABXD
XBCX
AXCD
When there are 2 out of 3 equal characters, the majority will win, so the result would be ABCD.
However, that's not so easy with different string sizes. If I expect a given size M (if scanning a price table, the rows are typically XX.XX), I can always penalize on strings bigger than M.
If we were talking numbers, a median filtering would work quite well (simple background subtraction in computer vision), or some least mean squares adaptive filtering.
There's also the problem of similar characters: l and 1 can be very similar, depending on the font.
I was also thinking of using string distances between each string. For example, choose the string with the smallest sum of distances with the others.
Has anyone addressed this kind of problem before? Is there any known algorithm for this kind of problem that I should know?
This problem is called multiple sequence alignment and you can read about it here
Related
I am using word2vec (and doc2vec) to get embeddings for sentences, but i want to completely ignore word order.
I am currently using gensim, but can use other packages if necessary.
As an example, my text looks like this:
[
['apple', 'banana','carrot','dates', 'elderberry', ..., 'zucchini'],
['aluminium', 'brass','copper', ..., 'zinc'],
...
]
I intentionally want 'apple' to be considered as close to 'zucchini' as it is to 'banana' so I have set the window size to a very large number, say 1000.
I am aware of 2 problems that may arise with this.
Problem 1:
The window might roll in at the start of a sentence creating the following training pairs:
('apple', ('banana')), ('apple', ('banana', 'carrot')), ('apple', ('banana', 'carrot', 'date')) before it eventually gets to the correct ('apple', ('banana','carrot', ..., 'zucchini')).
This would seem to have the effect of making 'apple' closer to 'banana' than 'zucchini',
since their are so many more pairs containing 'apple' and 'banana' than there are pairs containing 'apple' and 'zucchini'.
Problem 2:
I heard that pairs are sampled with inverse proportion to the distance from the target word to the context word- This also causes an issue making nearby words more seem more connected than I want them to be.
Is there a way around problems 1 and 2?
Should I be using cbow as opposed to sgns? Are there any other hyperparameters that I should be aware of?
What is the best way to go about removing/ignoring the order in this case?
Thank you
I'm not sure what you mean by "Problem 1" - there's no "roll" or "wraparound" in the usual interpretation of a word2vec-style algorithm's window parameter. So I wouldn't worry about this.
Regarding "Problem 2", this factor can be essentially made negligible by the choice of a giant window value – say for example, a value one million times larger than your largest sentence. Then, any difference in how the algorithm treats the nearest-word and the 2nd-nearest-word is vanishingly tiny.
(More specifically, the way the gensim implementation – which copies the original Google word2vec.c in this respect – achieves a sort of distance-based weighting is actually via random dynamic shrinking of the actual window used. That is, for each visit during training to each target word, the effective window truly used is some random number from 1 to the user-specified window. By effectively using smaller windows much of the time, the nearer words have more influence – just without the cost of performing other scaling on the whole window's words every time. But in your case, with a giant window value, it will be incredibly rare for the effective-window to ever be smaller than your actual sentences. Thus every word will be included, equally, almost every time.)
All these considerations would be the same using SG or CBOW mode.
I believe a million-times-larger window will be adequate for your needs, for if for some reason it wasn't, another way to essentially cancel-out any nearness effects could be to ensure your corpus's items individual word-orders are re-shuffled between each time they're accessed as training data. That ensures any nearness advantages will be mixed evenly across all words – especially if each sentence is trained on many times. (In a large-enough corpus, perhaps even just a 1-time shuffle of each sentence would be enough. Then, over all examples of co-occurring words, the word co-occurrences would be sampled in the right proportions even with small windows.)
Other tips:
If your training data starts in some arranged order that clumps words/topics together, it can be beneficial to shuffle them into a random order instead. (It's better if the full variety of the data is interleaved, rather than presented in runs of many similar examples.)
When your data isn't true natural-language data (with its usual distributions & ordering significance), it may be worth it to search further from the usual defaults to find optimal metaparameters. This goes for negative, sample, & especially ns_exponent. (One paper has suggested the optimal ns_exponent for training vectors for recommendation-systems is far different from the usual 0.75 default for natural-language modeling.)
I'm currently having a problem with the conception of an algorithm.
I want to create a WYSIWYG editor that goes along the current [bbcode] editor I have.
To do that, I use a div with contenteditable set to true for the WYSIWYG editor and a textarea containing the associated bbcode. Until there, no problem. But my concern is that if a user wants to add a tag (for example, the [b] tag), I need to know where they want to include it.
For that, I need to know exactly where in the bbcode I should insert the tags. I thought of comparing the two texts (one with html tags like <span>, the other with bbcode tags like [b]), and that's where I'm struggling.
I did some research but couldn't find anything that would help me, or I did not understand it correctly (maybe did I do a wrong research). What I could find is the Jaccard index, but I don't really know how to make it work correctly.
I also thought of another alternative. I could just take the code in the WYSIWYG editor before the cursor location, and split it every time I encounter a html tag. That way, I can, in the bbcode editor, search for the first occurrence, then search for the second occurrence starting at the last index found, and so on until I reach the place where the cursor is pointing at.
I'm not sure if it would work, and I find that solution a bit dirty. Am I totally wrong or should I do it this way?
Thanks for the help.
A popular way of determining what is the level of the similarity between the two texts is computing the mentioned Jaccard similarity. Citing Wikipedia:
The Jaccard index, also known as Intersection over Union and the Jaccard similarity coefficient, is a statistic used for comparing the similarity and diversity of sample sets. The Jaccard coefficient measures the similarity between finite sample sets, and is defined as the size of the intersection divided by the size of the union of the sample sets:
If you have a large number of texts though, computing the full Jaccard index of every possible combination of two texts is super computationally expensive. There is another way to approximate this index that is called minhashing. What it does is use several (e.g. 100) independent hash functions to create a signature and it repeats this procedure many times. This whole process has a nice property that the probability (over all permutations) that T1 = T2 is the same as J(A,B).
Another way to cluster similar texts (or any other data) together is to use Locality Sensitive Hashing which by itself is an approximation of what KNN does, and is usually worse than that, but is definitely faster to compute. The basic idea is to project the data into low-dimensional binary space (that is, each data point is mapped to a N-bit vector, the hash key). Each hash function h must satisfy the sensitive hashing property prob[h(x)=h(y)]=sim(x,y) where sim(x,y) in [0,1] is the similarity function of interest. For dots products it can be visualized as follows:
we can now ask what would be the has of the indicated point (in this case it's 101) and everything that is close to this point has the same hash.
EDIT to answer the comment
No, you asked about the text similarity and so I answered that. You basically ask how can you predict the position of the character in text 2. It depends on whether you analyze the writer's style or just pure syntax. In any of those two cases, IMHO you need some sort of statistics that will tell where it is likely for this character to occur given all the other data/text. You can go with n-grams, RNNs, LSTMs, Markov Chains or any other form of sequential data analysis.
Say, I have a number of strings which are quite similar but no absolutely identical.
They can differ more or less, but similarity can be seen by the naked eye.
All lengths are equal, each is 256 bytes. The total number of strings is less than 2^16.
What would be the best compression method for such case?
UPDATE (data format):
I can't share the data but I can describe it quite close to reality:
Imagine the notation (like LOGO language) which is the sequence of commands for some device for moving and drawing on plane. Such as:
U12 - move up 12 steps
D64 - move down 64 steps
C78 - change drawing color to 78
P1 - pen down (start drawing)
and so on.
The whole vocabulary of this language doesn't exceed the size of English alphabet.
The string then describes a whole picture: "U12C6P1L74D74R74U74P0....".
Imagine now the class of ten thousand children who were told to draw some very specific image with the help of this language: like the flag of their country. We will get 10K of strings which are all different and all alike at the same time.
Our task is compress the whole bunch of strings as good as possible.
My suspicion here is that there is a way to exploit this similarity and common length of the strings, while, Huffman e.g. wont use it explicitly.
Could you tell us what's the data ? Maybe like a DNA sequence ? Like
AGCTGTGCGAGAGAGAGCGGTGGG...
GGCTGTGCGAGCGAGAGCGGTGGG...
CGCTGTGAGAGNGAGAGCGGTGGG...
NGCTGTGCGAGAGAGAGCGGTGGG...
GGCTGTGCGAGTGAGAGCGGTGGG...
... ...
?
Maybe or not. Anyway here is two levels or two ways to think:
Huffman coding : ref. wikipedia by yourself
Stringology : ref. http://books.google.com.hk/books/about/Jewels_of_stringology.html?id=9NdohJXtIyYC
I think it's easy to solve your problem but hard to choose the best way. You can design several method to compare by using http://en.wikipedia.org/wiki/Data_compression and more tools .
Since you have a fix width of 256 bytes and it's a power of 2 I would try a burrow-wheeler transformation or a move-to-front algorithm with that size or maybe the double of that size. Then you can try a huffman code. Maybe you can try a hilbert curve on 256 bytes and then a bwt and mft?
"The total number of strings is less than 2^16." This is a small, bounded number, which makes your job very easy: Why don't you keep a lookup table (hash table) of all strings previously seen. You can then convert every line of 256 bytes into a two-byte index into this lookup table.
You then have a sequence of 16-bit integers. These integers will contains patterns like "after the pen went down, there is a 90% chance that the next command is to start to draw". If the data contains patterns like this, PPM is your choice. 7-zip has a high-quality PPM-implementation. You can choose it using the GUI or cmd-line.
I am in the process of writing a diff text tool to compare two similar source code files.
There are many such "diff" tools around, but mine shall be a little improved:
If it finds a set of lines are mismatched on both sides (ie. in both files), it shall not only highlight those lines but also highlight the individual changes in these lines (I call this inter-line comparison here).
An example of my somewhat working solution:
alt text http://files.tempel.org/tmp/diff_example.png
What it currently does is to take a set of mismatched lines and running their single chars thru the diff algo once more, producing the pink highlighting.
However, the second set of mismatches, containing "original 2", requires more work: Here, the first two right lines ("added line a/b") were added, while the third line is an altered version of the left side. I wish my software to detect this difference between a likely alteration and a probable new line.
When looking at this simple example, I can rather easily detect this case:
With an algo such as Levenshtein, I could find that of all right lines in the set of 3 to 5, line 5 matches left line 3 best, thus I could deduct that lines 3 and 4 on the right were added, and perform the inter-line comparison on left line 3 and right line 5.
So far, so good. But I am still stuck with how to turn this into a more general algorithm for this purpose.
In a more complex situation, a set of different lines could have added lines on both sides, with a few closely matching lines in between. This gets quite complicated:
I'd have to match not only the first line on the left to the best on the right, but vice versa as well, and so on with all other lines. Basically, I have to match every line on the left against every one on the right. At worst, this might create even crossings, so that it's not easily clear any more which lines were newly inserted and which were just altered (Note: I do not want to deal with possible moved lines in such a block, unless that would actually simplify the algorithm).
Sure, this is never going to be perfect, but I'm trying to get it better than it's now. Any suggestions that aren't too theoerical but rather practical (I'm not good understanding abstract algos) are appreciated.
Update
I must admit that I do not even understand how the LCS algo works. I simply feed it two arrays of strings and out comes a list of which sequences do not match. I am basically using the code from here: http://www.incava.org/projects/java/java-diff
Looking at the code I find one function equal() that is responsible for telling the algorithm whether two lines match or not. Based on what Pavel suggested, I wonder if that's the place where I'd make the changes. But how? This function only returns a boolean - not a relative value that could identify the quality of the match. And I can not simply used a fixed Levenshtein ration that would decide whether a similar line is still considered equal or not - I'll need something that's self-adopting to the entire set of lines in question.
So, what I'm basically saying is that I still do not understand where I'd apply the fuzzy value that relates to the relative similarity of lines that do not (exactly) match.
Levenshtein distance is based on the notion of an "edit script" that transforms one string into another. It's very closely related to the Needleman-Wunsch algorithm used for aligning DNA sequences by inserting gap characters, in which we search for the alignment that maximises a score in O(nm) time using dynamic programming. Exact matches between characters increase the score, while mismatches or inserted gap characters reduce the score. An example alignment of AACTTGCCA and AATGCGAT:
AACTTGCCA-
AA-T-GCGAT
(6 matches, 1 mismatch, 3 gap characters, 3 gap regions)
We can think of the top string being the "starting" sequence that we are transforming into the "final" sequence on the bottom. Each column containing a - gap character on the bottom is a deletion, each column with a - on the top is an insertion, and each column with different (non-gap) characters is a substitution. There are 2 deletions, 1 insertion and 1 substitution in the above alignment, so the Levenshtein distance is 4.
Here is another alignment of the same strings, with the same Levenshtein distance:
AACTTGCCA-
AA--TGCGAT
(6 matches, 1 mismatch, 3 gap characters, 2 gap regions)
But notice that although there are the same number of gaps, there is one less gap region. Because biological processes are more likely to create wide gaps than multiple separate gaps, biologists prefer this alignment -- and so will the users of your program. This is accomplished by also penalising the number of gap regions in the scores that we compute. An O(nm) algorithm to accomplish this for strings of lengths n and m was given by Gotoh in 1982 in a paper called "An improved algorithm for matching biological sequences". Unfortunately, I can't find any links to free full text of the paper -- but there are many useful tutorials that you can find by googling "sequence alignment" and "affine gap penalty".
In general, different choices of match, mismatch, gap and gap region weights will give different alignments, but any negative score for gap regions will prefer the bottom alignment above to the top one.
What does all this have to do with your problem? If you use Gotoh's algorithm on individual characters with a suitable gap penalty (arrived at with a few empirical tests), you should find a significant decrease in the the number of terrible-looking alignments like the example you gave.
Efficiency Considerations
Ideally, you could just do this on characters and ignore lines altogether, since the affine penalty will work to cluster changes into blocks spanning many lines wherever it can. But because of the higher running time, it may be more realistic to do a first pass on lines and then rerun the algorithm on characters, using as input all lines that are not identical. Under this scheme, any shared block of identical lines can be handled by compressing it into a single "character" with inflated matching weight, which helps to ensure no "crossings" appear.
With an algo such as Levenshtein, I could find that of all right lines in the set of 3 to 5, line 5 matches left line 3 best, thus I could deduct that lines 3 and 4 on the right were added, and perform the inter-line comparison on left line 3 and right line 5.
After you have determined it, use the same algorithm to determine what lines in these two chinks match each other. But you need to make slight modificaiton. When you used the algorithm to match equal lines, the lines could either match or not match, so that added either 0 or 1 to the cell of the table you used.
When comparing strings in one chunk some of them are "more equal" than others (ack. to Orwell). So they can add a real number from 0 to 1 to the cell when considering what sequence matches best so far.
To compute this metrics (from 0 to 1), you can apply to each pair of strings you encounter... right, the same algorithm again (actually, you already did this when you were doing the first pass of Levenstein algorithm). This will compute the length of LCS, whose ratio to the average length of two strings would be the the metric value.
Or, you can borrow the algorithm from one of diff tools. For instance, vimdiff can highlight the matches you require.
Here's one possible solution someone else just made me realize:
My original approach was like this:
Split the text up into separate lines and use LCS algo to determine where there are blocks of nonmatching lines.
Use some smart algo (which this question is about) to figure out which of these lines closely match, i.e. to tell that these lines were modified between revisions.
Compare those closely matching lines line-by-line using LCS again, while marking the non-matching lines as entirely new.
While this would allow for a better visual display of changes when comparing source code revisions, I now found that a much simpler approach is usually sufficient. It works like this:
Same as above.
Take the right and left block of nonmatching lines, concatenate those lines, and tokenize them (either into language-specific tokens/words, or just into single characters)
Apply the LCS algo on the two arrays of tokens.
Maybe those who replied to my original question assumed that I knew to do this all the time, but I had my focus so strongly on a per-line comparison that it did not occur to me to apply LCS on the set of lines by concatenating them, instead of processing them line-by-line.
So, while this approach will not provide as detailed change information as my original intent was, it still does improve the results over what I started yesterday with when I wrote this question.
I'll leave this question open for a while longer - maybe someone else, reading all this, can still provide a complete answer (Pavel and random_hacker offered some suggestions, but it's not a complete solution yet - anyway, thank you for the helpful comments).
Here at work, we often need to find a string from the list of strings that is the closest match to some other input string. Currently, we are using Needleman-Wunsch algorithm. The algorithm often returns a lot of false-positives (if we set the minimum-score too low), sometimes it doesn't find a match when it should (when the minimum-score is too high) and, most of the times, we need to check the results by hand. We thought we should try other alternatives.
Do you have any experiences with the algorithms?
Do you know how the algorithms compare to one another?
I'd really appreciate some advice.
PS: We're coding in C#, but you shouldn't care about it - I'm asking about the algorithms in general.
Oh, I'm sorry I forgot to mention that.
No, we're not using it to match duplicate data. We have a list of strings that we are looking for - we call it search-list. And then we need to process texts from various sources (like RSS feeds, web-sites, forums, etc.) - we extract parts of those texts (there are entire sets of rules for that, but that's irrelevant) and we need to match those against the search-list. If the string matches one of the strings in search-list - we need to do some further processing of the thing (which is also irrelevant).
We can not perform the normal comparison, because the strings extracted from the outside sources, most of the times, include some extra words etc.
Anyway, it's not for duplicate detection.
OK, Needleman-Wunsch(NW) is a classic end-to-end ("global") aligner from the bioinformatics literature. It was long ago available as "align" and "align0" in the FASTA package. The difference was that the "0" version wasn't as biased about avoiding end-gapping, which often allowed favoring high-quality internal matches easier. Smith-Waterman, I suspect you're aware, is a local aligner and is the original basis of BLAST. FASTA had it's own local aligner as well that was slightly different. All of these are essentially heuristic methods for estimating Levenshtein distance relevant to a scoring metric for individual character pairs (in bioinformatics, often given by Dayhoff/"PAM", Henikoff&Henikoff, or other matrices and usually replaced with something simpler and more reasonably reflective of replacements in linguistic word morphology when applied to natural language).
Let's not be precious about labels: Levenshtein distance, as referenced in practice at least, is basically edit distance and you have to estimate it because it's not feasible to compute it generally, and it's expensive to compute exactly even in interesting special cases: the water gets deep quick there, and thus we have heuristic methods of long and good repute.
Now as to your own problem: several years ago, I had to check the accuracy of short DNA reads against reference sequence known to be correct and I came up with something I called "anchored alignments".
The idea is to take your reference string set and "digest" it by finding all locations where a given N-character substring occurs. Choose N so that the table you build is not too big but also so that substrings of length N are not too common. For small alphabets like DNA bases, it's possible to come up with a perfect hash on strings of N characters and make a table and chain the matches in a linked list from each bin. The list entries must identify the sequence and start position of the substring that maps to the bin in whose list they occur. These are "anchors" in the list of strings to be searched at which an NW alignment is likely to be useful.
When processing a query string, you take the N characters starting at some offset K in the query string, hash them, look up their bin, and if the list for that bin is nonempty then you go through all the list records and perform alignments between the query string and the search string referenced in the record. When doing these alignments, you line up the query string and the search string at the anchor and extract a substring of the search string that is the same length as the query string and which contains that anchor at the same offset, K.
If you choose a long enough anchor length N, and a reasonable set of values of offset K (they can be spread across the query string or be restricted to low offsets) you should get a subset of possible alignments and often will get clearer winners. Typically you will want to use the less end-biased align0-like NW aligner.
This method tries to boost NW a bit by restricting it's input and this has a performance gain because you do less alignments and they are more often between similar sequences. Another good thing to do with your NW aligner is to allow it to give up after some amount or length of gapping occurs to cut costs, especially if you know you're not going to see or be interested in middling-quality matches.
Finally, this method was used on a system with small alphabets, with K restricted to the first 100 or so positions in the query string and with search strings much larger than the queries (the DNA reads were around 1000 bases and the search strings were on the order of 10000, so I was looking for approximate substring matches justified by an estimate of edit distance specifically). Adapting this methodology to natural language will require some careful thought: you lose on alphabet size but you gain if your query strings and search strings are of similar length.
Either way, allowing more than one anchor from different ends of the query string to be used simultaneously might be helpful in further filtering data fed to NW. If you do this, be prepared to possibly send overlapping strings each containing one of the two anchors to the aligner and then reconcile the alignments... or possibly further modify NW to emphasize keeping your anchors mostly intact during an alignment using penalty modification during the algorithm's execution.
Hope this is helpful or at least interesting.
Related to the Levenstein distance: you might wish to normalize it by dividing the result with the length of the longer string, so that you always get a number between 0 and 1 and so that you can compare the distance of pair of strings in a meaningful way (the expression L(A, B) > L(A, C) - for example - is meaningless unless you normalize the distance).
We are using the Levenshtein distance method to check for duplicate customers in our database. It works quite well.
Alternative algorithms to look at are agrep (Wikipedia entry on agrep),
FASTA and BLAST biological sequence matching algorithms. These are special cases of approximate string matching, also in the Stony Brook algorithm repositry. If you can specify the ways the strings differ from each other, you could probably focus on a tailored algorithm. For example, aspell uses some variant of "soundslike" (soundex-metaphone) distance in combination with a "keyboard" distance to accomodate bad spellers and bad typers alike.
Use FM Index with Backtracking, similar to the one in Bowtie fuzzy aligner
In order to minimize mismatches due to slight variations or errors in spelling, I've used the Metaphone algorithm, then Levenshtein distance (scaled to 0-100 as a percentage match) on the Metaphone encodings for a measure of closeness. That seems to have worked fairly well.
To expand on Cd-MaN's answer, it sounds like you're facing a normalization problem. It isn't obvious how to handle scores between alignments with varying lengths.
Given what you are interested in, you may want to obtain p-values for your alignment. If you are using Needleman-Wunsch, you can obtain these p-values using Karlin-Altschul statistics http://www.ncbi.nlm.nih.gov/BLAST/tutorial/Altschul-1.html
BLAST will can local alignment and evaluate them using these statistics. If you are concerned about speed, this would be a good tool to use.
Another option is to use HMMER. HMMER uses Profile Hidden Markov Models to align sequences. Personally, I think this is a more powerful approach since it also provides positional information. http://hmmer.janelia.org/