package owl

  1. Overview
  2. Docs
Legend:
Library
Module
Module type
Parameter
Class
Class type

NLP: Corpus module

Type definition
type t

Type of a text corpus.

Query corpus
val length : t -> int

Return the size of the corpus, i.e. number of documents.

val get : t -> int -> string

Return the ith document in the corpus.

val get_tok : t -> int -> int array

Return the ith tokenised document in the corpus.

val get_uri : t -> string

Return the path of the corpus.

val get_bin_uri : t -> string

Return the path of the binary format of corpus.

val get_bin_fh : t -> Stdlib.in_channel

Return the file handle of the binary formation of corpus.

val get_tok_uri : t -> string

Return the path of tokenised corpus.

val get_tok_fh : t -> Stdlib.in_channel

Return the file handle of the tokenised corpus.

val get_vocab_uri : t -> string

Return the path of vocabulary file associated with the corpus.

val get_vocab : t -> Owl_nlp_vocabulary.t

Return the vocabulary associated with the corpus.

val get_docid : t -> int array

Return a list of document ids which are mapped back to the original file where the corpus is built.

Iteration functions
val next : t -> string

Return the next document in the corpus.

val next_tok : t -> int array

Return the next tokenised document in the corpus.

val iteri : (int -> string -> unit) -> t -> unit

Iterate all the documents in the corpus, the index (line number) is passed in.

val iteri_tok : (int -> int array -> unit) -> t -> unit

Iterate the tokenised documents in the corpus, the index (line number) is passed in.

val mapi : (int -> string -> 'a) -> t -> 'a array

Map all the documents in a corpus into another array. The index (line number) is passed in.

val mapi_tok : (int -> 'a -> 'b) -> t -> 'b array

Map all the tokenised ocuments in a corpus into another array. The index (line number) is passed in.

val next_batch : ?size:int -> t -> string array

Return the next batch of documents in a corpus as a string array. The default ``size`` is 100.

val next_batch_tok : ?size:int -> t -> int array array

Return the next batch of tokenised documents in a corpus as a string array. The default ``size`` is 100.

val reset_iterators : t -> unit

Reset the iterator to the beginning of the corpus.

Core functions
val build : ?docid:int array -> ?stopwords:(string, 'a) Stdlib.Hashtbl.t -> ?lo:float -> ?hi:float -> ?vocab:Owl_nlp_vocabulary.t -> ?minlen:int -> string -> t

This function builds up a corpus of type ``t`` from a given raw text corpus. We assume that each line in the raw text corpus represents a document.

Parameters: * ``?docid``: passed in ``docid`` can be used for tracking back to the original corpus, but this is not compulsory. * ``?stopwords``: stopwords used in building vocabulary. * ``?lo``: any word below this lower bound of the frequency is removed from vocabulary. * ``?hi``: any word above this upper bound of the frequency is removed from vocabulary. * ``?vocab``: an optional vocabulary, if it is not passed, the vocabulary is built from current corpus. * ``?(minlen=10)``: threshold of the document length, any document shorter than this is removed from the corpus. * ``fname``: the file name of the raw text corpus.

val tokenise : t -> string -> int array

``tokenise corpus doc`` tokenises the document ``doc`` using the ``corpus`` and its associated vocabulary.

val unique : string -> string -> int array

Remove the duplicates in a text corpus, the ids of the removed files are returned.

val simple_process : string -> string

Function for simple pre-processing a given string.

val preprocess : (string -> bytes) -> string -> string -> unit

``preprocess f input_file output_file`` pre-processes a given file ``input_file`` with the passed in function ``f`` then saves the output to ``output_file``.

E.g., you can plug in ``simple_process`` function to clean up the text. Note this function will not change the number of lines in a corpus.

I/O functions
val save : t -> string -> unit

Serialise the corpus and save it to a file of given name.

val load : string -> t

Load a serialised corpus from a file.

val save_txt : t -> string -> unit

Convert the tokenised corpus back to a text file

val to_string : t -> string

The string representation of a corpus, contains the summary of a corpus.

val print : t -> unit

Pretty print the summary of a text corpus.

Helper functions
val create : string -> int array -> int array -> Stdlib.in_channel option -> Stdlib.in_channel option -> Owl_nlp_vocabulary.t option -> int -> int array -> t

```create uri bin_ofs tok_ofs bin_fh tok_fh vocab minlen docid` wraps up the corpus into a record of type ``t``.

val reduce_model : t -> t

Set some fields to ``None`` so it can be safely serialised.

val cleanup : t -> unit

Close the opened file handles associated with the corpus.

OCaml

Innovation. Community. Security.