The LAPPS Grid project, which has developed a platform providing access to a vast array of language processing tools and resources for the purposes of research and development in natural language processing (NLP), has recently expanded to enhance its usability by non-technical users such as those in the DH community. We provide a live demonstration of LAPPS Grid use, ranging from “from scratch" construction of a workflow using atomic tools to a pre-configured docker image that can be run off-the-shelf on a laptop or in the cloud, for several tasks of relevance to the DH community.
Over the past few years, Digital Humanities (DH) has looked to Computational Linguistics (CL) for methods to enable richer analysis of literary, historical, and other kinds of documents, recognizing that CL methods and procedures can in fact enhance the kinds and amount of information that can be automatically extracted from language data [14]. However, several obstacles have prevented humanists from wholesale adoption of CL tools, the most well known of which is that they are typically difficult to use without a fair amount of technical background. Other, more subtle but perhaps more deeply rooted obstacles have also contributed, most notably dramatic differences in perspective, approach, and simply differences in the language data that each community typically deals with. It is only recently that CL methods and tools have begun to be made more accessible to non-technical users and are beginning to be widely adopted by the DH community; however, there remains considerable work to be done to fully adapt CL tools and methods to use by DH scholars.
The Language Applications (LAPPS) Grid [6] is an NSF-funded project involving Vassar College, Brandeis University, Carnegie Mellon University, and the Linguistic Data Consortium at the University of Pennsylvania. The original motivation for the project, begun in 2012, was to address the endemic lack of interoperability among CL tools and data that has plagued the CL field for decades. Atomic natural language processing (NLP) tools (e.g., part of speech taggers, syntactic analyzers, entity detectors, etc.) are typically pipelined to create more sophisticated applications; the lack of interoperability among tools, corpora, and other language resources often leads to considerable waste of effort to make them work together in a pipeline, or “workflow". To overcome the problem, the LAPPS Grid project undertook to engineer a platform that both provides access to a wide array of language processing tools and resources, and exploits recognized standards and best practices to negotiate incompatibilities for the user.
Over the past five years the LAPPS Grid project has collaborated with several major projects in the US, Europe, and Asia to expand its range of accessible tools and resources as well as to augment the capabilities of the platform. Our collaborators serve a broad range of users, well beyond the NLP community we originally intended to serve, including users involved in inter-cultural communication and users from the DH community. We have also begun to create purpose-built instances of the LAPPS Grid to use in courses aimed at non-technical users, and we are currently working with a major project in the digital humanities [12] and pursuing funding to collaborate with several others. As a result, the LAPPS Grid is continually increasing its usability by non-technical users such as those in the DH community.
Our demonstration provides several sample usages of the LAPPS Grid relevant to digital humanities research, including tool pipelines developed “from scratch" as well as pre-configured workflows that can be used as is, and demonstrates both the analysis and creation of resources. 2
The LAPPS Grid is an open platform that provides access to hundreds of NLP
tools and language resources. It incorporates the Galaxy workflow and data
management framework [
The LAPPS Grid is part of the Federated Grid of Language Services (FGLS)
[
The LAPPS Grid in its current form addresses many of the needs for DH research. It provides easy-to-use access to a wide variety of customizable low-level CL tools, including tokenizers, sentence boundary detectors, part-of-speech taggers, named entity recognizers, co-referencers, phrase-structure and dependency parsers, among others. It also provides facilities for comparing the effectiveness of tools that perform the same task in order to identify the one that is best suited to the task. For example, Figure 1 shows an evaluation pipeline in Galaxy that compares the output of three named entity recognizers to from gold standard annotations; this example shows each small step in the workflow, but sub-steps (for 1http://langrid.org 2http://weblicht.sfs.uni-tuebingen.de/ 3https://lindat.mff.cuni.cz/ example, the Tokenization-SentenceSplitter-Tagger sequence that feeds the three entity recognizers) could be bundled into a workflow and plugged in as a single step.
The datasets used in DH research are diverse, often involving ancient text, texts
in languages typically not covered in NLP such as Latin, poetry, historical
documents, and multi-media, and in some cases need to be representative across
multiple genres. Large CL datasets, on the other hand, are typically largely composed
of genres such as newswire (Penn Treebank [
Data visualization is often essential for humanities research, and the LAPPS
Grid includes a wide range of statistical and visualization tools. A basic but
common task is to generate frequency distributions or distributions across a text,
collection, timeline, etc. for any type of phenomenon. For example, one recent study
examined the appearance of neologisms and words that become obsolete over
several decades of Dutch magazine texts as well as tweets, by generating graphs
showing initial and final word frequencies over time intervals [13]. Other projects use
visualization of relations in graph form. For example, one study used named entity
recognition and co-reference tools to identify characters in the novels
comprising A Song of Ice and Fire and then generated a weighted graph depicting social
relations among characters based on dialogue interactions [
The LAPPS Grid demonstration will show how it can be used to perform tasks
relevant to DH research such as those described above, as well as many others.
Facilities suitable for DH scholarship and research not currently available in the
LAPPS Grid are being regularly added to the platform as we receive input from
the DH community, and our current collaboration with the CLARIN projects in
Europe will significantly enhance LAPPS Grid facilities for DH research in the
near future. In the meantime, LAPPS Grid users already have access to the wide
range of tools and resources available through the Language Grid and other
federated grids, which focus on machine translation and other facilities for cultural
collaboration. A new collaboration with the Alveo project [
European Language Resources Association (ELRA). [14] Christopher Welty and Nancy Ide. Using the right tools: Enhancing retrieval from marked-up documents. Computers and the Humanities, 33(1-2):59–84, 1999.