Textmatch

Find fuzzy matches between datasets.

Fuzzy matching is the art and science of connecting up bits of information that are written differently but represent the same thing – such as a person or company.

Installing

$ pip install textmatch

Getting started

The best way to approach fuzzy matching with Textmatch is to start with an exact match. From there, you can incrementally improve the results by telling Textmatch about relevant information that should be taken into account and irrelevant information that should be disregarded. Experiment with different approaches. It is helpful to know what the data looks like, and how it has been collected.

The input datasets can be dataframes from PyArrow, Pandas, or Polars. The output results will be in PyArrow format – which can then be converted to Pandas with matches.to_pandas(), or to Polars with polars.from_arrow(matches).

Example

import textmatch

data1:

name	place	codename
Sam Collins	Vietnam	none
Roy Bland	London	Soldier
George Smiley	London	Beggerman
Bill Haydon	London	Tailor
Perçy Allélíne	London	Tinker
Kretzschmar	Hamburg	none
Oliver Lacon	London	none
Jim Prideaux	Slovakia	none
Peter Guillam Esq	Brixton	none
Toby Esterhase	Vienna	Poorman
Connie Sachs	Oxford	none

data2:

Person Name	Location
Maria Andreyevna Ostrakova	Russia
Konny Saks	Oxford
Tony Esterhase	Vienna
Peter Guillam	Brixton
Mr Jim Prideaux	Czech Republic
Lacon Oliver	Cambridge
Claus Kretzschmar	Hamburg
Richard Bland	London
Roy Rodgers	Romania
Percy Alleline	London
Bill-Haydon	London
George SMILEY	London
Roy Bland	UK
Sam Collins	Vietnam

To run an exact match on the name column from the first dataset against Person Name from the second:

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']]
)

The resulting matches include the two names which are written exactly the same:

name	place	codename	Person Name	Location
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

Matches are many-to-many, ie. it is possible for one row in the first dataset to match several rows in the second, and vice-versa.

Tip

There is a tradeoff between false negatives and false positives – it is often better to have some incorrect matches in your results that can be manually checked afterwards than to have correct matches missing.

Usage

Textmatch has one function, run, which accepts the first dataset followed by the second. All other arguments are optional, described below:

Fields

The fields1 and fields2 arguments accept two-dimensional lists of column names that should be used for the match. These should be in the same order for both datasets – the first column specified for the first dataset will be compared against the first column specified for the second dataset, and so on. The columns can only contain strings. Defaults to comparing all columns. The outer list is used for blocking.

Example

To match on the name and place columns from the first dataset against Person Name and Location from the second:

textmatch.run(
  data1,
  data2,
  fields1=[['name', 'place']],
  fields2=[['Person Name', 'Location']]
)

The resulting matches include the single name-place pair which is the same in both datasets:

name	place	codename	Person Name	Location
Sam Collins	Vietnam	none	Sam Collins	Vietnam

Ignorance

The ignores argument accepts a two-dimensional list of characteristics which should be disregarded for two records to be considered a match. The outer list is used for blocking.

Combining different forms of ignorance can be quite powerful. The order in which you specify them is not significant.

case ignores how text is capitalised.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['case']]
)

The resulting matches include George Smiley, whose surname is in all-capitals in the second dataset:

name	place	codename	Person Name	Location
George Smiley	London	Beggerman	George SMILEY	London
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

nonalpha ignores anything that isn't a number or a letter.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['nonalpha']]
)

The resulting matches include Bill Haydon, whose name is written with a hypen in the second dataset:

name	place	codename	Person Name	Location
Bill Haydon	London	Tailor	Bill-Haydon	London
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

nonlatin ignores non-Latin characters – so Jérôme will match Jerome, Weiß will match Weiss, and Пушкин will match Pushkin. The further the script is from the Latin alphabet, the less accurate this transliteration will be.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['nonlatin']]
)

The resulting matches include Percy Alleline, whose name is written with several diacritics in the first dataset:

name	place	codename	Person Name	Location
Perçy Allélíne	London	Tinker	Percy Alleline	London
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

words-leading ignores all words except the last. This is useful for matching on surnames only.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['words-leading']]
)

The resulting matches include Toby and Tony Esterhase, Jim Prideaux and Mr Jim Prideaux, Kretzschmar and Claus Kretzschmar, as well as Roy and Richard Bland:

name	place	codename	Person Name	Location
Toby Esterhase	Vienna	Poorman	Tony Esterhase	Vienna
Jim Prideaux	Slovakia	none	Mr Jim Prideaux	Czech Republic
Kretzschmar	Hamburg	none	Claus Kretzschmar	Hamburg
Roy Bland	London	Soldier	Richard Bland	London
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

words-tailing ignore all words except the first.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['words-tailing']]
)

The resulting matches include Peter Guillam Esq and Peter Guillam, Roy Bland and Roy Rodgers, as well as the two capitalisations of George Smiley:

name	place	codename	Person Name	Location
Peter Guillam Esq	Brixton	none	Peter Guillam	Brixton
Roy Bland	London	Soldier	Roy Rodgers	Romania
George Smiley	London	Beggerman	George SMILEY	London
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

words-order ignores the order in which the words are given. This is useful for matching names given surname-first with those given forename-first.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['words-order']]
)

The resulting matches include Oliver Lacon, whose name is written surname-first in the second dataset:

name	place	codename	Person Name	Location
Oliver Lacon	London	none	Lacon Oliver	Cambridge
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

titles ignores common English name prefixes such as Mr and Ms. There is a full list of these titles here.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['titles']]
)

The resulting matches include Jim Prideaux, who has the title 'Mr' in the second dataset:

name	place	codename	Person Name	Location
Jim Prideaux	Slovakia	none	Mr Jim Prideaux	Czech Republic
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

regex ignores terms specific to your data using a given regular expression. This is specified inline: regex=EXPRESSION.

Example

To use the regular expression Esq$ to ignore the word 'Esq' where it appear at the end of a value:

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['regex= Esq$']]
)

The resulting matches include Peter Guillam, who has the name suffix 'Esq' in the first dataset:

name	place	codename	Person Name	Location
Peter Guillam Esq	Brixton	none	Peter Guillam	Brixton
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

Methods & thresholds

The methods argument accepts a list of methods. This lets you specify the algorithm which is used to do the matching. Subsequent items in the list are used for blocking.

There are three different categories of method:

• Compared methods work by comparing every row from the first dataset with every row from the second, producing a number that represents the degree of the match. This means the amount of time required to run a match grows exponentially with the size of the input datasets. However, they are still useful for larger matches when using blocking.
• Applied methods transform text into a different representation before they are matched up. These methods are quicker than compared ones, though no meaningful matching degree number is generated – either they match or they don't.
• Custom methods have their own individual approach. Textmatch only has one, Bilenko. It generates a matching degree number.

For those matching methods that generate a matching degree number there is then a threshold filter for any two records to be considered to be a match – you can adjust this with the threshold argument, which accepts a list of numbers between 0.0 and 1.0, defaulting to 0.6. Subsequent items in the list are also used for blocking.

You can also include the matching degree number as a column by specifying it in the outputs.

Warning

When working with names of people, exact matches, even when other pieces of information such as birthdays are included, are not a guarantee that the two names actually refer to the same human. Furthermore, the chance of a mismatch is unintuitively high – as illustrated by the birthday paradox.

literal is the default – it matches strings exactly, after ignored characteristics have been taken into account.

levenshtein Uses the Damerau-Levenshtein string distance metric that counts the number of changes that would have to be made to transform one string into another. Performs compared matching. Where two strings are of different lengths the longer string is used as the denominator for the threshold filter. Good at picking up typos and other small differences in spelling.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  methods=['levenshtein']
)

The resulting matches include various names with small typographical differences, though the most emblematic of this matching method would be Toby and Tony Esterhase:

name	place	codename	Person Name	Location
Sam Collins	Vietnam	none	Sam Collins	Vietnam
Roy Bland	London	Soldier	Roy Bland	UK
George Smiley	London	Beggerman	George SMILEY	London
Bill Haydon	London	Tailor	Bill-Haydon	London
Perçy Allélíne	London	Tinker	Percy Alleline	London
Kretzschmar	Hamburg	none	Claus Kretzschmar	Hamburg
Jim Prideaux	Slovakia	none	Mr Jim Prideaux	Czech Republic
Peter Guillam Esq	Brixton	none	Peter Guillam	Brixton
Toby Esterhase	Vienna	Poorman	Tony Esterhase	Vienna

jaro uses the Jaro-Winkler string distance metric that counts characters in common, though it considers differences near the start of the string to be more significant than differences near the end. Performs compared matching. It tends to work better than Levenshtein for shorter strings of text.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  methods=['jaro']
)

The resulting matches includes many more matches than levenshtein, though also many more false positives:

name	place	codename	Person Name	Location
Sam Collins	Vietnam	none	Percy Alleline	London
Sam Collins	Vietnam	none	Sam Collins	Vietnam
Roy Bland	London	Soldier	Richard Bland	London
Roy Bland	London	Soldier	Roy Rodgers	Romania
Roy Bland	London	Soldier	Bill-Haydon	London
Roy Bland	London	Soldier	Roy Bland	UK
George Smiley	London	Beggerman	George SMILEY	London
Bill Haydon	London	Tailor	Bill-Haydon	London
Bill Haydon	London	Tailor	Roy Bland	UK
Perçy Allélíne	London	Tinker	Peter Guillam	Brixton
Perçy Allélíne	London	Tinker	Percy Alleline	London
Kretzschmar	Hamburg	none	Claus Kretzschmar	Hamburg
Jim Prideaux	Slovakia	none	Mr Jim Prideaux	Czech Republic
Peter Guillam Esq	Brixton	none	Peter Guillam	Brixton
Toby Esterhase	Vienna	Poorman	Tony Esterhase	Vienna
Toby Esterhase	Vienna	Poorman	Roy Rodgers	Romania
Connie Sachs	Oxford	none	Konny Saks	Oxford

metaphone uses the Double Metaphone phonetic encoding algorithm to convert words into a representation of how they are pronounced. Performs applied matching. Tends to work better for data which has been transcribed or transliterated.

Example

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  methods=['metaphone']
)

The resulting matches includes those with nonalphanumeric differences, as well as Connie Sachs and Konny Saks, names written quite differently that would be pronounced the same:

name	place	codename	Person Name	Location
Connie Sachs	Oxford	none	Konny Saks	Oxford
Roy Bland	London	Soldier	Roy Bland	UK
George Smiley	London	Beggerman	George SMILEY	London
Sam Collins	Vietnam	none	Sam Collins	Vietnam
Bill Haydon	London	Tailor	Bill-Haydon	London

bilenko uses Dedupe, a library built by Forest Gregg and Derek Eder based on the work of Mikhail Bilenko that will ask you to train it by asking whether different pairs of records should match. The information you give it is then extrapolated to match up the rest of the dataset. The more examples you give it, the better the results will be. At minimum, try to provide 10 positive matches and 10 negative matches. Performs custom matching.

Blocking

Blocking is the approach of performing multiple matches, with subsequent matches only applying to the subset of matches resulting from the previous match. This can make matches both quicker and more precise. This is an advanced topic, and can be ignored if you are happy with the quality of matches and are dealing with smaller datasets.

In a 'regular' match, you are really just matching using a single block. Each block is defined by: a list of fields for each dataset, a list of ignores, a method, and a threshold. To perform a blocked match Textmatche needs to know each of these things for each block. You specify these through list arguments, or through outer lists for those arguments where the block requires a list itself. If you specify one of these things less than the total number of blocks – such as if you had two blocks, but specified the threshold once – that value will then be used for all subsequent blocks.

Example

To specify a first block that does a case-insensitive literal match on surnames, then a second block performing a Levenshtein match on forenames:

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  ignores=[['case', 'words-leading'], ['words-tailing']],
  methods=['literal', 'levenshtein']
)

name	place	codename	Person Name	Location
Kretzschmar	Hamburg	none	Claus Kretzschmar	Hamburg
George Smiley	London	Beggerman	George SMILEY	London
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam

Outputs

The output argument accepts a list of column names which should appear in the output, each prefixed with a number and a dot indicating which dataset that field is from. They are case-sensitive, and can be in any order you desire. It defaults to all columns in the first dataset, followed by all columns in the second.

There are some special column definitions: 1* and 2* expand into all columns from the first and second datasets respectively, and degree will add a column with the matching degree number.

Example

To include every column from the second dataset, followed by the codename column from the first, followed by the matching degree:

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  methods=['levenshtein'],
  output=['2*', '1.codename', 'degree']
)

Person Name	Location	codename	degree
Sam Collins	Vietnam	none	1.0
Roy Bland	UK	Soldier	1.0
George SMILEY	London	Beggerman	0.61538464
Bill-Haydon	London	Tailor	0.90909094
Percy Alleline	London	Tinker	0.78571427
Claus Kretzschmar	Hamburg	none	0.64705884
Mr Jim Prideaux	Czech Republic	none	0.8
Peter Guillam	Brixton	none	0.7647059
Tony Esterhase	Vienna	Poorman	0.9285714

Join types

The join argument takes a string that indicates what other nonmatching records should be included in the output. A left-outer join will return everything from the first dataset, whether there was a match or not, a right-outer to do the same but for the second dataset, and a full-outer to return everything from both datasets. Where two rows didn't match the values will be blank. Defaults to an inner join, where only successful matches are returned.

Example

To include all rows from the first dataset, but only those that match from the second:

textmatch.run(
  data1,
  data2,
  fields1=[['name']],
  fields2=[['Person Name']],
  join='left-outer'
)

name	place	codename	Person Name	Location
Roy Bland	London	Soldier	Roy Bland	UK
Sam Collins	Vietnam	none	Sam Collins	Vietnam
George Smiley	London	Beggerman
Bill Haydon	London	Tailor
Perçy Allélíne	London	Tinker
Kretzschmar	Hamburg	none
Oliver Lacon	London	none
Jim Prideaux	Slovakia	none
Peter Guillam Esq	Brixton	none
Toby Esterhase	Vienna	Poorman
Connie Sachs	Oxford	none

Progress bars & alerts

By default Textmatch does not print out any details of its operations, however it is possible to to display progress bars and logging alerts by defining progress and alert functions that handle these events. This is especially useful in an interactive Jupyter environment.

For example, using tqdm and ipywidgets:

import tqdm.notebook

def progress(operation, total):
    bar = tqdm.notebook.tqdm(desc=operation, total=total, bar_format='{desc} {bar} {percentage:3.0f}% {remaining} left', dynamic_ncols=True)
    return bar.update

def alert(message, importance):
    print(f'[{importance.upper()}] {message}')

These functions are then passed as arguments when you run Textmatch:

textmatch.run(
    data1,
    data2,
    fields1=[['name']],
    fields2=[['Person Name']],
    progress=progress,
    alert=alert
).to_pandas()

Other libraries

Python

• RecordLinkage
• Splink
• DeezyMatch
• Dedupe
• DeepMatcher
• Hlink
• Skrub
• RLTK
• String Grouper
• PyEntityMatching
• TextPack
• String2String
• RapidFuzz
• TheFuzz

R

• Zoomerjoin
• Reclin2
• Fedmatch
• FastLink
• Fuzzyjoin