Quid is a tool for quotation detection in texts and can deal with common properties of quotations, for example, ellipses or inaccurate quotations.
Project description
Readme
Quid is a tool for quotation detection in texts and can deal with common properties of quotations, for example, ellipses or inaccurate quotations.
Overview
Quid is a tool to find quotations in two texts, called source and target. If known, the source text should be the one that is quoted by the target text. This allows the algorithm to handle things like ellipsis in quotations, e.g.
0 52 This is a long Text and the long test goes on and on
0 45 This is a long Text [...] test goes on and on
Installation
pip install Quid
Usage
There are two ways to use the algorithm. The following two sections describe the use of the algorithm in code and from the command line.
In code
The algorithm can be found in the package quid
. To use it create a Quid
object which can be configured with the following arguments:
- The minimum number of tokens of a match (default: 5)
- The maximum number of tokens to skip when extending a match backwards (default: 10)
- The maximum number of tokens to skip when extending a match forwards (default: 3)
- The maximum distance in tokens between to matches considered for merging (default: 2)
- The maximum distance in tokens between two matches considered for merging where the target text contains an ellipses between the matches (default: 10)
- Whether to include matched text in the returned data structure (default: True)
- How to handle ambiguous matches. If False, for a match with multiple matched segments in the source text, multiple matches will be returned. Otherwise, only the first match will be returned. (default: False)
- The threshold for the minimal levenshtein similarity between tokens (and the initial n-grams) to be accepted as a match (default: 0.85)
- Whether to split texts which are longer than the threshold (in words) defined with
split_length
for faster processing (default: False) - The threshold for splitting texts (in number of words) (default: 30000)
- The maximum number of processes for parallel processing (default: 1)
Then call the compare
method on the object which expects two texts to be compared.
The method returns a list with the following structure: List[Match]
. Match
stores two MatchSpans
. One for the source text and one for the target text. MatchSpan
stores the start
and end
character positions for the matching spans in the source and target text.
from quid.core.Quid import Quid
quid = Quid()
matches = quid.compare('file 1 content', 'file 2 content')
Command line
The quid compare
command provides a command line interface to the algorithm.
usage: quid compare [-h] [--text] [--no-text]
[--output-type {json,text,csv}] [--csv-sep CSV_SEP]
[--output-folder-path OUTPUT_FOLDER_PATH]
[--min-match-length MIN_MATCH_LENGTH]
[--look-back-limit LOOK_BACK_LIMIT]
[--look-ahead-limit LOOK_AHEAD_LIMIT]
[--max-merge-distance MAX_MERGE_DISTANCE]
[--max-merge-ellipsis-distance MAX_MERGE_ELLIPSIS_DISTANCE]
[--create-dated-subfolder]
[--no-create-dated-subfolder]
[--max-num-processes MAX_NUM_PROCESSES]
[--keep-ambiguous-matches]
[--no-keep-ambiguous-matches]
[--min-levenshtein-similarity MIN_LEVENSHTEIN_SIMILARITY]
[--split-long-texts] [--no-split-long-texts]
[--split-length SPLIT_LENGTH]
source-file-path target-path
Quid compare allows the user to find quotations in two texts, a source text
and a target text. If known, the source text should be the one that is quoted
by the target text. This allows the algorithm to handle things like ellipsis
in quotations.
positional arguments:
source-file-path Path to the source text file
target-path Path to the target text file or folder
optional arguments:
-h, --help show this help message and exit
--text Include matched text in the returned data structure
--no-text Don't include matched text in the returned data
structure
--output-type {json,text,csv}
The output type
--csv-sep CSV_SEP output separator for csv (default: '\t')
--output-folder-path OUTPUT_FOLDER_PATH
The output folder path. If this option is set the
output will be saved to a file created in the
specified folder
--min-match-length MIN_MATCH_LENGTH
The minimum number of tokens of a match (>= 1,
default: 5)
--look-back-limit LOOK_BACK_LIMIT
The maximum number of tokens to skip when extending a
match backwards (>= 0, default: 10)
--look-ahead-limit LOOK_AHEAD_LIMIT
The maximum number of tokens to skip when extending a
match forwards (>= 0, default: 3)
--max-merge-distance MAX_MERGE_DISTANCE
The maximum distance in tokens between two matches
considered for merging (>= 0, default: 2)
--max-merge-ellipsis-distance MAX_MERGE_ELLIPSIS_DISTANCE
The maximum distance in tokens between two matches
considered for merging where the target text contains
an ellipsis between the matches (>= 0, default: 10)
--create-dated-subfolder
Create a subfolder named with the current date to
store the results
--no-create-dated-subfolder
Don't create a subfolder named with the current date
to store the results
--max-num-processes MAX_NUM_PROCESSES
Maximum number of processes to use for parallel
processing
--keep-ambiguous-matches
For a match with multiple matched segments in the
source text, multiple matches will be returned.
--no-keep-ambiguous-matches
For a match with multiple matched segments in the
source text, only the first match will be returned.
--min-levenshtein-similarity MIN_LEVENSHTEIN_SIMILARITY
The threshold for the minimal levenshtein similarity
between tokens (and the initial n-grams) to be
accepted as a match (between 0 and 1, default: 0.85)
--split-long-texts Split texts longer than split-length words for
fasterprocessing
--no-split-long-texts
Do not split texts longer than 30000 tokens.
--split-length SPLIT_LENGTH
If split-long-texts is set to True, texts longer (in
number of words) than this threshold will be split for
faster processing.
By default, the result is returned as a json structure: List[Match]
. Match
stores two MatchSpans
. One for the source text and one for the target text. MatchSpan
stores the start
and end
character positions for the matching spans in the source and target text.
For example,
[
{
"source_span": {
"start": 0,
"end": 52,
"text": "This is a long Text and the long test goes on and on"
},
"target_span": {
"start": 0,
"end": 45,
"text": "This is a long Text [...] test goes on and on"
}
}
]
Alternatively, the result can be printed in a human-readable text format, e.g.:
0 52 This is a long Text and the long test goes on and on
0 45 This is a long Text [...] test goes on and on
In case the matching text is not needed, the option --no-text allows to exclude the text from the output.
Processing "long" texts
Depending on the length of the texts and the hardware used, processing times can get quite long. For texts longer than
a couple of hundreds of thousands characters, it can make sense to use the --split-long-texts
command line option (or
split_long_texts
argument) and set --max-num-processes
(or max_num_processes
argument) to define the number of
parallel processes to be used. If --split-long-texts
is used, texts longer than the default of 30000 tokens will be
split. This limit can also be changed using the --split-length
command line option (or split_length
argument).
When run from the command line, using --split-long-texts
automatically shows a progress bar. To show a progress bar
when using Quid in code, the show_progress
argument can be set to True
.
Note: --split-long-texts
does not work in combination with comparing multiple target texts (i.e. passing a folder as
target-path
).
Passager
The package passager
contains code to extract key passages from the found matches. The passage
command produces several json files.
The resulting data structure is documented in the data structure readme.
Usage
usage: quid passage [-h]
source-file-path target-folder-path
matches-folder-path output-folder-path
Quid passage allows the user to extract key passages from the found
matches.
positional arguments:
source-file-path Path to the source text file
target-folder-path Path to the target texts folder path
matches-folder-path Path to the folder with the match files
output-folder-path Path to the output folder
Visualization
The package visualization
contains code to create the content for a web page to visualize the key passages.
For a white label version of the website, see QuidEx-wh.
Usage
usage: quid visualize [-h] [--title TITLE] [--author AUTHOR]
[--year YEAR] [--censor]
source-file-path target-folder-path
passages-folder-path output-folder-path
Quid visualize allows the user to create the files needed for a website that
visualizes the Quid algorithm results.
positional arguments:
source-file-path Path to the source text file
target-folder-path Path to the target texts folder path
passages-folder-path
Path to the folder with the key passages files, i.e.
the resulting files from Quid passage
output-folder-path Path to the output folder
optional arguments:
-h, --help show this help message and exit
--title TITLE Title of the work
--author AUTHOR Author of the work
--year YEAR Year of the work
Logging
By default, the log level is set to WARN
. This can be changed with the --log-level
command line option.
For example:
quid --log-level INFO compare …
Performance
For in-depth information on the evaluation, see our paper below. Perfomance of the current version of Quid is as follows:
Work | Precision | Recall | F-Score |
---|---|---|---|
Die Judenbuche | 0.82 | 0.93 | 0.87 |
Micheal Kohlhaas | 0.71 | 0.93 | 0.80 |
History
Quid was formerly known as Lotte and later renamed. Earlier publications use the name Lotte.
Citation
If you use Quid or base your work on our code, please cite our paper:
@inproceedings{arnold2021lotte,
title = {{L}otte and {A}nnette: {A} {F}ramework for {F}inding and {E}xploring {K}ey {P}assages in {L}iterary {W}orks},
author = {Arnold, Frederik and Jäschke, Robert},
booktitle = {Proceedings of the Workshop on Natural Language Processing for Digital Humanities},
year = {2021},
publisher = {NLP Association of India (NLPAI)},
url = {https://aclanthology.org/2021.nlp4dh-1.7},
pages = {55--63}
}
Acknowledgements
The algorithm is inspired by sim_text by Dick Grune ^1 and Similarity texter: A text-comparison web tool based on the “sim_text” algorithm by Sofia Kalaidopoulou (2016) ^2
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.