bhlnames

bhlnames takes one or more scientific names and finds their occurrences in Biodiversity Heritage Library. The app also creates an apparent link between a name-string/publication input and the corresponding BHL reference.

Introduction

Biodiversity Heritage Library (BHL) contains more than 250 000 volumes (books, scientific journals, diaries of explorers, etc.). BHL provides an important biodiversity information. Since the middle of 18th century, scientists use Latinized scientific names as identifiers for known species. For biologists, it is crucial to get information about species in publications, it is especially valuable to find original description of species or creation of a new binomial genus/species combination.

This program tries to answer the following questions:

• Where a particular name-string appears in BHL?

• Given a names string, where species it assigned to appears in BHL? In this case we find all synonyms of the name-string and the currently accepted name of the species

• Given a name-string and a reference to its original publication, does this publication exists in BHL, and what is the link to it?

• Given a name-string and a reference to a publication about the name-string, find a link to this publication in BHL.

The bhlnames app uses Catalogue Of Life (CoL) synonymy information to find publications not only about a given name but also about its synonyms. In the future it will support other resources with synonymy information.

Provided functionalities

1. Find references in BHL where a scientific name-string appears.

   Searching for a name-string without synonyms.

2. Find references in BHL where a taxon, represented by a name-string appears.

   Determining the corresponding taxon an entered name-string. This taxon has a currently accepted name and a variety of synonym name-strings. Entered name-string might match either a synonym, or the currently accepted name. Searching for references where any of the found name-strings appear.

3. Find a reference to the original description of a name or a new combination.

   In this case the input is a nomenclatural event represented by a name and its official nomenclatural publication. We use provided information trying to find a BHL reference that corresponds to that publication.

4. Find a link to a publication, using a publication reference and a name used in the publication.

5. Provide REST API and command line tool to access aforementioned functionalities.

Prerequisites

To be able to use this program you need

• a modern computer (laptop or desktop)
• one of the 3 operating systems (Linux, Mac OS, Windows)
• a PostgreSQL database
• 30+ GB of space on a hard drive
• 32GB or more of memory

Installation

1. Download the latest release of bhlnames, untar or unzip the executable bhlnames or bhlnames.exe and place it somewhere in your PATH.

2. Create a database (default name is bhlnames) on your Postgresql server. We are not covering how to use Postgresql in this document. There are many tutorials about it on the web. Make sure that the database is accessible from the computer where you installed bhlnames executable.

3. Run bhlnames -V. It will show the version number of the program and will create a configuration file. Terminal output will provide information where the file is located. For example on Linux machines it will be placed in $HOME/.config/bhlnames.yaml.

4. Go to the configuration file and modify database parameters in the config. You can also change setup for InputDir directory for downloaded and temporary files, as well as for key-value store databases. You can leave other parameters as is for now.

The system should be ready for the initialization step.

Initialization

This step downloads all the needed BHL and names metadata on your computer. Some of the data go to the Postgresql database, others to a key-value store. You do not need to worry about the creation of tables, or key-value databases; they will be populated automatically. The program uses a file containing metadata dump from BHL, as well as names data created by bhlindex. Dump provides information about papers (parts in BHL terminology), volumes (issues), and books/journal volume sets (titles). The bhlindex provides fresh information about names and their occurrences in BHL.

To start the initialization process type:

bhlnames init

The whole process will take about 1-2 hours, but it could take significantly longer if your computer or internet connection are slow.

If for some reason you have to restart the program, you do not need to delete working directories or the database. All of them will be updated automatically. Some slow steps will not be repeated (such as downloading full dump of BHL database), unless you use -r option during initialization:

bhlnames init -r

In this case, everything will start from the beginning. In case if you downloaded a newer version of bhlnames instead of older ones, you might need to drop the database and create it again, and do init from scratch. Note that the BHL dump is updated regularly, and it is good to rebuild your metadata set from time to time from scratch.

Usage

To find references to a whole taxon (synonyms and currently accepted name) from a name-string:

bhlnames name "Pardosa moesta"

The result (in JSON format) will be sent to STDOUT and can be redirected to a file

bhlnames name "Pardosa moesta" > pm.json

By default, JSON is returned in its compact form. Optionally you can get data in a more human-readable form with:

bhlnames name "Pardosa moesta" -f pretty

You can also use jq or a similar tool

bhlnames name "Pardosa moesta" 2>/dev/null | jq

In case if it is preferable to have the oldest publications last, you can reverse sorting order with:

bhlnames name "Pardosa moesta" -f pretty -d

To search for a large collection of names provide the name of a file instead (one name per line):

bhlnames name names.txt

For computers with modern multi-core CPU, you can increase number of parallel jobs. Usually, there is no much gain from setting more than 8 jobs.

bhlnames name names.txt -j 8

To get a short version of data without details for references:

bhlnames name names.txt -s

To get results without synonyms:

bhlnames name names.txt --no_synonyms

To find a link to a name-string with its original reference you can use a CSV.

Mimumal fields are:

Id,NameString,RefString

Where Id is an internal identifier, NameString is a full scientific name with authorship and year (if they are given), RefString is an unparsed reference string.

If you have a more detailed information the following fields can also be used:

NameCanonical canonical form of a name, NameYear the year of the name publication, RefYear the year of the reference publication.

Most

You can use the following command:

bhlnames nomen name-refs.csv

The result will be send to STDOUT in a compact JSON format, one datum per line. Use jq or similar program to render 'pretty' version of JSON.

On a 12-core laptop, processing of 10000 names took about 40 seconds with 8 parallel jobs, and 2m 45sec with a single job. 10000 names generated 120MB of results.

When you find the optimal number of jobs for your computer you can modify JobsNum parameter in your version of the bhlnames.yaml file accordingly.

REST API

To start bhlnames as a server on a port 1234:

bhlnames rest -p 1234

REST end-points

• /name_refs (POST) to find occurrences of a names-string. Takes JSON encoded list of name-strings as an argument.

• /taxon_refs (POST) to find occurrences of a taxon. Takes JSON encoded list of name-strings as an argument.

• /nomen_refs (POST) to find a link to the provided reference. Takes a JSON-encoded structure.

For more details how to use API you can refer to the REST test file.

Explanation of received data

Taxon and name-string output

This output is created by bhlnames name command.

Returning information can be quite large. You will get data in chronological order. If there is data about a specific paper, we return information about the paper; if we only have information about the item, or title, we return that information.

For every item, we return the most populated Linnean kingdom for the unique names found in the item. We also provide the percentage of names that got resolved to that kingdom. We use the managerial CoL classification for this purpose.

In addition, we return a "context" of the item. The context is the lowest taxon that still contains at least 50% of all names found in the item. This gives a better idea about the item's biological content. For example, if the context is "Araneae" the item is mostly about spiders.

You can use provided page_id and item_id to find information on BHL website. For example to find page_id 26895127 use:

https://www.biodiversitylibrary.org/page/26895127

Original reference output

This output is created by bhlnames nomen command.

You receive input information (id, name, reference) as well as 0 or 1 reference from BHL that was the picked as the best candidate to the link for the input reference. If we did not get any feasible candidates, no BHL reference is provided.

TODO: Explain score and score thresholds.

Development

Running tests

1. Install direnv, copy envrc.example to .envrc.

2. Modify .envrc according to your needs, enable direnv.

3. Install bhlnames as described above.

4. Use bhlnames test database for data.

5. in ./internal/io/restio/restio.go go to TODO and swap development and production servers. (Dont forget to swap them back for production)

6. Start RESTful service in another terminal window:

   bhlnames rest

7. Run make tools

8. Run make test

9. Run make testrest

Authors

• Dmitry Mozzherin, Geoff Ower

License

Released under MIT license