There is function to parse SequenceExample --> tf.parse_single_sequence_example().
But it parses only single SequenceExample, which is not effective.
Is there any possibility to parse a batch of SequenceExamples?
tf.parse_example can parse many Examples.
Documentation for tf.parse_example contain a little info about SequenceExample:
Each FixedLenSequenceFeature df maps to a Tensor of the specified type (or tf.float32 if not specified) and shape (serialized.size(), None) + df.shape. All examples in serialized will be padded with default_value along the second dimension.
But it is not clear, how to do that. Have not found any examples in google.
Is it possible to parse many SequenceExamples using parse_example() or may be other function exists?
Edit:
Where can I ask question to tensorflow developers: does they plan to implement parse function for multiple SequenceExample -s?
Any help ll be appreciated.
If you have many small sequences where batching at this stage is important, I would recommend VarLenFeatures or FixedLenSequenceFeatures with regular Example protos (which, as you note, can be parsed in batches with parse_example). For examples of this, see the unit tests associated with example parsing (testSerializedContainingSparse parses Examples with FixedLenSequenceFeatures).
SequenceExamples are more geared toward cases where there is significant amounts of preprocessing work to be done for each SequenceExample (which can be done in parallel with queues). parse_example does does not support SequenceExamples.
Related
I'm trying to parse a very large file using FParsec. The file's size is 61GB, which is too big to hold in RAM, so I'd like to generate a sequence of results (i.e. seq<'Result>), rather than a list, if possible. Can this be done with FParsec? (I've come up with a jerry-rigged implementation that actually does this, but it doesn't work well in practice due to the O(n) performance of CharStream.Seek.)
The file is line-oriented (one record per line), which should make it possible in theory to parse in batches of, say, 1000 records at a time. The FParsec "Tips and tricks" section says:
If you’re dealing with large input files or very slow parsers, it
might also be worth trying to parse multiple sections within a single
file in parallel. For this to be efficient there must be a fast way to
find the start and end points of such sections. For example, if you
are parsing a large serialized data structure, the format might allow
you to easily skip over segments within the file, so that you can chop
up the input into multiple independent parts that can be parsed in
parallel. Another example could be a programming languages whose
grammar makes it easy to skip over a complete class or function
definition, e.g. by finding the closing brace or by interpreting the
indentation. In this case it might be worth not to parse the
definitions directly when they are encountered, but instead to skip
over them, push their text content into a queue and then to process
that queue in parallel.
This sounds perfect for me: I'd like to pre-parse each batch of records into a queue, and then finish parsing them in parallel later. However, I don't know how to accomplish this with the FParsec API. How can I create such a queue without using up all my RAM?
FWIW, the file I'm trying to parse is here if anyone wants to give it a try with me. :)
The "obvious" thing that comes to mind, would be pre-processing the file using something like File.ReadLines and then parsing one line at a time.
If this doesn't work (your PDF looked, like a record is a few lines long), then you can make a seq of records or 1000 records or something like that using normal FileStream reading. This would not need to know details of the record, but it would be convenient, if you can at least delimit the records.
Either way, you end up with a lazy seq that the parser can then read.
I'm making an application using a dependency tree parser. Actually, the parser is this one:
Parser Stanford, but it rarely change one or two letters of some words in a sentence that I want to parse. This is a big trouble for me, because I can't see any pattern in these changes and I need the dependency tree with the same words of my sentence.
All I can see is that just some words have these problems. I'm working with a tweets database. So, I have a lot of grammar mistakes in this data. For example the hashtag '#AllAmericanhumour ' becomes AllAmericanhumor. It misses one letter(u).
Is there anything I can do to solve this problem? In my first view I thought using an edit distance algorithm, but I think that might be an easier way to do it.
Thanks everybody in advance
You can give options to the tokenizer with the -tokenize.options flag/property. For this particular normalization, you can turn it off with
-tokenize.options americanize=false
There are also various other normalizations that you can turn off (see PTBTokenizer or http://nlp.stanford.edu/software/tokenizer.shtml. You can turn off a lot with
-tokenize.options ptb3Escaping=false
However, the parser is trained on data that looks like the output of ptb3Escaping=true and so will tend to degrade in performance if used with unnormalized tokens. So, you may want to consider alternative strategies.
If you're working at the Java level, you can look at the word tokens, which are actually Maps, and they have various keys. OriginalTextAnnotation will give you the unnormalized token, even when it has been normalized. CharacterOffsetBeginAnnotation and CharacterOffsetEndAnnotation will map to character offsets into the text.
p.s. And you should accept some answers :-).
I am searching for information on algorithms to process text sentences or to follow a structure when creating sentences that are valid in a normal human language such as English. I would like to know if there are projects working in this field that I can go learn from or start using.
For example, if I gave a program a noun, provided it with a thesaurus (for related words) and part-of-speech (so it understood where each word belonged in a sentence) - could it create a random, valid sentence?
I'm sure there are many sub-sections of this kind of research so any leads into this would be great.
The field you're looking for is called natural language generation, a subfield of natural language processing
http://en.wikipedia.org/wiki/Natural_language_processing
Sentence generation is either really easy or really hard depending on how good you want the sentences to be. Currently, there aren't programs that will be able to generate 100% sensible sentences about given nouns (even with a thesaurus) -- if that is what you mean.
If, on the other hand, you would be satisfied with nonsense that was sometimes ungrammatical, then you could try an n-gram based sentence generator. These just chain together of words that tend to appear in sequence, and 3-4-gram generators look quite okay sometimes (although you'll recognize them as what generates a lot of spam email).
Here's an intro to the basics of n-gram based generation, using NLTK:
http://www.nltk.org/book/ch02.html#generating-random-text-with-bigrams
This is called NLG (Natural Language Generation), although that is mainly the task of generating text that describes a set of data. There is also a lot of research on completely random sentence generation as well.
One starting point is to use Markov chains to generate sentences. How this is done is that you have a transition matrix that says how likely it is to transition between every every part-of-speech. You also have the most likely starting and ending part-of-speech of a sentence. Put this all together and you can generate likely sequences of parts-of-speech.
Now, you are far from done, this will first of all not offer a very good result as you are only considering the probability between adjacent words (also called bi-grams), so what you want to do is to extend this to look for instance at the transition matrix between three parts-of-speech (this makes a 3D matrix and gives you trigrams). You can extend it to 4-grams, 5-grams, etc. depending on the processing power and if your corpus can fill such matrix.
Lastly, you need to patch up things such as object agreement (subject-verb-agreement, adjective-verb-agreement (not in English though), etc.) and tense, so that everything is congruent.
Yes. There is some work dealing with solving problems in NLG with AI techniques. As far as I know, currently, there is no method that you can use for any practical use.
If you have the background, I suggest getting familiar with some work by Alexander Koller from Saarland University. He describes how to code NLG to PDDL. The main article you'll want to read is "Sentence generating as a planning problem".
If you do not have any background in NLP, just search for the online courses or course materials by Michael Collings or Dan Jurafsky.
Writing random sentences is not that hard. Any parser textbook's simple-english-grammar example can be run in reverse to generate grammatically correct nonsense sentences.
Another way is the word-tuple-random-walk, made popular by the old BYTE magazine TRAVESTY, or stuff like
http://www.perlmonks.org/index.pl?node_id=94856
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 6 years ago.
Improve this question
I am a college student getting my Computer Science degree. A lot of my fellow students really haven't done a lot of programming. They've done their class assignments, but let's be honest here those questions don't really teach you how to program.
I have had several other students ask me questions about how to parse things, and I'm never quite sure how to explain it to them. Is it best to start just going line by line looking for substrings, or just give them the more complicated lecture about using proper lexical analysis, etc. to create tokens, use BNF, and all of that other stuff? They never quite understand it when I try to explain it.
What's the best approach to explain this without confusing them or discouraging them from actually trying.
I'd explain parsing as the process of turning some kind of data into another kind of data.
In practice, for me this is almost always turning a string, or binary data, into a data structure inside my Program.
For example, turning
":Nick!User#Host PRIVMSG #channel :Hello!"
into (C)
struct irc_line {
char *nick;
char *user;
char *host;
char *command;
char **arguments;
char *message;
} sample = { "Nick", "User", "Host", "PRIVMSG", { "#channel" }, "Hello!" }
Parsing is the process of analyzing text made of a sequence of tokens to determine its grammatical structure with respect to a given (more or less) formal grammar.
The parser then builds a data structure based on the tokens. This data structure can then be used by a compiler, interpreter or translator to create an executable program or library.
(source: wikimedia.org)
If I gave you an english sentence, and asked you to break down the sentence into its parts of speech (nouns, verbs, etc.), you would be parsing the sentence.
That's the simplest explanation of parsing I can think of.
That said, parsing is a non-trivial computational problem. You have to start with simple examples, and work your way up to the more complex.
What is parsing?
In computer science, parsing is the process of analysing text to determine if it belongs to a specific language or not (i.e. is syntactically valid for that language's grammar). It is an informal name for the syntactic analysis process.
For example, suppose the language a^n b^n (which means same number of characters A followed by the same number of characters B). A parser for that language would accept AABB input and reject the AAAB input. That is what a parser does.
In addition, during this process a data structure could be created for further processing. In my previous example, it could, for instance, to store the AA and BB in two separate stacks.
Anything that happens after it, like giving meaning to AA or BB, or transform it in something else, is not parsing. Giving meaning to parts of an input sequence of tokens is called semantic analysis.
What isn't parsing?
Parsing is not transform one thing into another. Transforming A into B, is, in essence, what a compiler does. Compiling takes several steps, parsing is only one of them.
Parsing is not extracting meaning from a text. That is semantic analysis, a step of the compiling process.
What is the simplest way to understand it?
I think the best way for understanding the parsing concept is to begin with the simpler concepts. The simplest one in language processing subject is the finite automaton. It is a formalism to parsing regular languages, such as regular expressions.
It is very simple, you have an input, a set of states and a set of transitions. Consider the following language built over the alphabet { A, B }, L = { w | w starts with 'AA' or 'BB' as substring }. The automaton below represents a possible parser for that language whose all valid words starts with 'AA' or 'BB'.
A-->(q1)--A-->(qf)
/
(q0)
\
B-->(q2)--B-->(qf)
It is a very simple parser for that language. You start at (q0), the initial state, then you read a symbol from the input, if it is A then you move to (q1) state, otherwise (it is a B, remember the remember the alphabet is only A and B) you move to (q2) state and so on. If you reach (qf) state, then the input was accepted.
As it is visual, you only need a pencil and a piece of paper to explain what a parser is to anyone, including a child. I think the simplicity is what makes the automata the most suitable way to teaching language processing concepts, such as parsing.
Finally, being a Computer Science student, you will study such concepts in-deep at theoretical computer science classes such as Formal Languages and Theory of Computation.
Have them try to write a program that can evaluate arbitrary simple arithmetic expressions. This is a simple problem to understand but as you start getting deeper into it a lot of basic parsing starts to make sense.
Parsing is about READING data in one format, so that you can use it to your needs.
I think you need to teach them to think like this. So, this is the simplest way I can think of to explain parsing for someone new to this concept.
Generally, we try to parse data one line at a time because generally it is easier for humans to think this way, dividing and conquering, and also easier to code.
We call field to every minimum undivisible data. Name is field, Age is another field, and Surname is another field. For example.
In a line, we can have various fields. In order to distinguish them, we can delimit fields by separators or by the maximum length assign to each field.
For example:
By separating fields by comma
Paul,20,Jones
Or by space (Name can have 20 letters max, age up to 3 digits, Jones up to 20 letters)
Paul 020Jones
Any of the before set of fields is called a record.
To separate between a delimited field record we need to delimit record. A dot will be enough (though you know you can apply CR/LF).
A list could be:
Michael,39,Jordan.Shaquille,40,O'neal.Lebron,24,James.
or with CR/LF
Michael,39,Jordan
Shaquille,40,O'neal
Lebron,24,James
You can say them to list 10 nba (or nlf) players they like. Then, they should type them according to a format. Then make a program to parse it and display each record. One group, can make list in a comma-separated format and a program to parse a list in a fixed size format, and viceversa.
Parsing to me is breaking down something into meaningful parts... using a definable or predefined known, common set of part "definitions".
For programming languages there would be keyword parts, usable punctuation sequences...
For pumpkin pie it might be something like the crust, filling and toppings.
For written languages there might be what a word is, a sentence, what a verb is...
For spoken languages it might be tone, volume, mood, implication, emotion, context
Syntax analysis (as well as common sense after all) would tell if what your are parsing is a pumpkinpie or a programming language. Does it have crust? well maybe it's pumpkin pudding or perhaps a spoken language !
One thing to note about parsing stuff is there are usually many ways to break things into parts.
For example you could break up a pumpkin pie by cutting it from the center to the edge or from the bottom to the top or with a scoop to get the filling out or by using a sledge hammer or eating it.
And how you parse things would determine if doing something with those parts will be easy or hard.
In the "computer languages" world, there are common ways to parse text source code. These common methods (algorithims) have titles or names. Search the Internet for common methods/names for ways to parse languages. Wikipedia can help in this regard.
In linguistics, to divide language into small components that can be analyzed. For example, parsing this sentence would involve dividing it into words and phrases and identifying the type of each component (e.g.,verb, adjective, or noun).
Parsing is a very important part of many computer science disciplines. For example, compilers must parse source code to be able to translate it into object code. Likewise, any application that processes complex commands must be able to parse the commands. This includes virtually all end-user applications.
Parsing is often divided into lexical analysis and semantic parsing. Lexical analysis concentrates on dividing strings into components, called tokens, based on punctuationand other keys. Semantic parsing then attempts to determine the meaning of the string.
http://www.webopedia.com/TERM/P/parse.html
Simple explanation: Parsing is breaking a block of data into smaller pieces (tokens) by following a set of rules (using delimiters for example),
so that this data could be processes piece by piece (managed, analysed, interpreted, transmitted, ets).
Examples: Many applications (like Spreadsheet programs) use CSV (Comma Separated Values) file format to import and export data. CSV format makes it possible for the applications to process this data with a help of a special parser.
Web browsers have special parsers for HTML and CSS files. JSON parsers exist. All special file formats must have some parsers designed specifically for them.
Parsec is designed to parse textual information, but it occurs to me that Parsec could also be suitable to do binary file format parsing for complex formats that involve conditional segments, out-of-order segments, etc.
Is there an ability to do this or a similar, alternative package that does this? If not, what is the best way in Haskell to parse binary file formats?
The key tools for parsing binary files are:
Data.Binary
cereal
attoparsec
Binary is the most general solution, Cereal can be great for limited data sizes, and attoparsec is perfectly fine for e.g. packet parsing. All of these are aimed at very high performance, unlike Parsec. There are many examples on hackage as well.
You might be interested in AttoParsec, which was designed for this purpose, I think.
I've used Data Binary successfully.
It works fine, though you might want to use Parsec 3, Attoparsec, or Iteratees. Parsec's reliance on String as its intermediate representation may bloat your memory footprint quite a bit, whereas the others can be configured to use ByteStrings.
Iteratees are particularly attractive because it is easier to ensure they won't hold onto the beginning of your input and can be fed chunks of data incrementally a they come available. This prevents you from having to read the entire input into memory in advance and lets you avoid other nasty workarounds like lazy IO.
The best approach depends on the format of the binary file.
Many binary formats are designed to make parsing easy (unlike text formats that are primarily to be read by humans). So any union data type will be preceded by a discriminator that tells you what type to expect, all fields are either fixed length or preceded by a length field, and so on. For this kind of data I would recommend Data.Binary; typically you create a matching Haskell data type for each type in the file, and then make each of those types an instance of Binary. Define the "get" method for reading; it returns a "Get" monad action which is basically a very simple parser. You will also need to define a "put" method.
On the other hand if your binary data doesn't fit into this kind of world then you will need attoparsec. I've never used that, so I can't comment further, but this blog post is very positive.