8 minute read

Guidelines for the Assignment: FINAL

This Mapping Assignment, otherwise known as Assignment 2, will be completed in one step. It builds on work we did in class, particularly with annotating “1, the Road” by Ross Goodwin with the web-based semantic annotation tool Recogito.

Manual geocoding is essentially a human identification and contextualization of placenames found in a text and matching them with possible geographic information using contextual clues. Automatic geocoding is based on a machine identifying possible place names through two predictive processes (1) assigning certain words the tag of geographic identity and (2) matching those against databases without contextualization.

About the Text

The text you will annotate in Recogito is “1 the Road,” the AI-created novel by Ross Goodwin. You have been assigned a section of the novel to annotate in class and you should finish it by the time you start this assignment. The remaining parts not assigned to you have been completed by your instructor.

Assignment Details

  • Format: Individual or pairs (maximum 2 people, it can be someone other than you annotated with)
  • Length: Approximately 1500 words (about an 8-minute read), plus visuals, including a clickable leaflet interactive map in your Github page.
  • Supplementary files: Located in the DHS F26 shared folder > data
  • Due Date: Friday, 7 November 2026, 11:59pm.

This assignment has three main elements:

  1. Manual Annotation with Recogito: You will be provided with a text in plain text format to crowd-annotate manually in the web-based platform Recogito that we used in class. You will also be provided with a reference pdf of 1 The Road as well as a .txt file. Each of the teams was given a section to annotate (remember that section 5 was repeated by accident). This annotation should be completed by the end of class on 27 October 2026.

  2. Supplementary Annotation: using any other digital resource–Google Maps, OSM, etc., you can complete (to whatever extent is possible) your geo-annotation of 1 The Road. My suggestion is that you work only on your assigned section and others work on theirs. When you are satisfied with what you have, use the download options to export the data.

  3. Comparison with Automated Geocoding: You will be provided with the automatically tagged text and structured geodata for this section. Assemble your evidence, analysis, and visuals in a web-published essay in the form of a post that tells a coherent story about your findings.

Step 1: Research the Text in Question

Before you begin analysis, research the texts themselves: Who is the author? What is the publication context? What are the general themes and contents of the novel? You can take notes on what your impressions of the novel are. The introductions to the book in the pdf version are helpful in this regard.

Step 2: Annotating in Recogito

This step is all done inside the platform Recogito. The txt file was uploaded and shared with you in Recogito (file name “1TheRoad_sectioned.txt”), and in class we marked up as much of the geographic information in your assigned section as you see fit. Feel free to take notes on what doesn’t work or what isn’t easy. Also take screenshots as necessary.

Consider what McDowell wrote in the introduction to 1 The Road (in the pdf version):

“it knew locations and businesses (like the Biloxi Hard Rock Cafe) that were proximate because they were exposed by the Foursquare API, which is to say, by the priorities of the techno-capitalist producer and subject. These locations were often gas stations and fast-food restaurants. There are many on the American road trip.” (p. 28)

Specify in your assignment which parts of the text you and your partner marked up.

Step 3: Downloading and Cleaning

Using download options download a csv version of your dataset and do any cleaning necessary. In particular, you may need to move any geodata that was entered manually from the comments field to the latlong fields. You can use whatever version of the data you have downloaded in csv and geojson formats.

Specific the download date for the file you used and discuss any cleaning you may have done. Once you have downloaded the data, you are done with Recogito.

When you have finished cleaning your csv file, you should place it in a new folder inside of your assets folder (in your Github repository, EDIT FOR CLARITY) for me to examine. Point to this folder with a link in your essay.

Step 4: Visualizing the Data

You will have three datasets to visualize in this step.

A. The first one is the one you generated above in Step 3a.

For the other two, there is no more annotation to do!

B. The second one has been automatically (code) generated using a rule based NER system. I have used a Python script for geographical entity recognition using a library called spaCy and geocoding with geopy that I used to do that for the “I, the Road” exercise. That file can be found in drive: A2 NER spaCy.

C. The third one has been generated with chatGPT Sol 5.6. chatGPT Sol 5.6 was instructed to create a TEI transcription of 1 the Road and identified geographic information such as place names and latitude/longitude coordinates. It used spaCy’s Named Entity Recognition (NER) to automatically find likely places in the text and tagged them with , while recognizable coordinates were tagged with . Finally, it used geopy to look up coordinates for place names that did not already have them, producing a table of places and their locations. You will find the file with the annotations `road_pages_41-140_recogito.xml` shared in Recogito. You can also find the XML file and the resulting CSV file in Drive: A2 LLM approach.

For step 4, you should use the notebook in posit.cloud entitled “Interactive Leaflet Map with Two Clickable Layers: 1 the Road”. The notebook employs the leaflet package in R to create a self-contained html map with three layers, similar to Assignment 1. It also performs a very slight distortion of the points on the map to prevent overlap (occultation) to make the next step easier.

Once you have saved the map as an html file to your satisfaction using the notebook, download it out of posit.cloud. Place it in the assets folder you used before for maps, then tnsert it into a post in your site as you did in Assignment 1 (refer to instructions there).

Step 5: Conduct your Analysis

In this last step of the exercise, you can bring together your notes from the above steps to compare the visual results and the structured data on which they depend. Your ultimate goal here is to compare the processes of manual and automatic geocoding (of two kinds) to explain what kinds of differ in human and machine approaches. You will probably find McDowell’s comments excerpted above helpful. Feel free to do any additional research to distinguish between the kinds of geocoding.

Guiding Questions (you do not need to answer all these questions):

Observations:

  • Do you observe any patterns when comparing the manually and automatically geocoded locations? between the two different automatically geocoded locations?
  • How did the different groups annotate differently.
  • Do the two methods produce similar coordinates or mappings?
  • Where do the largest discrepancies occur? At different scales? kinds of places?

Agreement/disagreement of the two methods:

  • Identify one or two cases where manual and automatic geocoding differ significantly. What is happening in those examples?
  • How different are the human annotations in different sections made by different annotators?
  • Are mismatches random, or do they follow patterns (e.g., ambiguous place names, incomplete data)?
  • Are certain types of locations (cities, regions, historical places, commercial locations) more prone to disagreement?
  • Optional: take a look at the NER processes with in the two scripts using spaCy. What are its assumptions about place names and place name extraction? What is the gazetteer used by geopy to get latitude/longitude pairs? Can you find out any information about it?

Data and Sensitivity:

  • How does the quality or completeness of the original place names affect each geocoding method?
  • What happens when place names are: misspelled, have OCR errors, are too granular or vague (e.g., “near Atlanta”).

Interpretation vs automation:

  • What kinds of decisions are humans making in manual geocoding that automated systems cannot easily replicate?
  • What assumptions does the automated system make when matching place names to coordinates?
  • Can you think of a case when automatic geocoding might have significant implications?
  • How does each method handle ambiguity (e.g., multiple places with the same name)?
  • Comment on the multiple levels of automation that take place here. Do they help us or hinder us in understanding “human space”?
  • How does color and/or jitter help make the map more visible?

Critical Questions:

  • What kinds of places or regions might be systematically misrepresented by automated geocoding systems?
  • How did Goodwin’s creation of data by “Wordcar”–including pulling information from the Foursquare API–create a certain kind of data?
  • What does this comparison reveal about the limits of computational approaches to spatial data?
  • In the video and introduction, they mention that the car produced many thousand words. How long is the novel? Can studying the map help us to understand anything about how much of the automatically created text was trimmed by the editor?

To include in your assignment

  • analysis of the process and product responding to some of the guiding questions above
  • captioned screenshots of relevant issues in Recogito, the RMarkdown notebook or the mapping
  • an iframe of the three layer map created with an appropriate captions and discussion of making it
  • mention of the quote above, any additional research, relevant information from Drucker’s chapter on mapping.

Group Work and AI

You should not need to use AI for any parts of this assignment. If you do, to look at the data or to expand the code or to make sense of the python scripts, explain what you did.

As in the previous assignment, give a few sentence breakdown of how you worked collectively, if you work in pairs.

Assessment

Your work will be assessed according to the following criteria located here.

Marking your work as complete

It is fine to publish your assignment iteratively, but when you finish the final version of your assignment, write at the bottom of it “READY FOR GRADING”.

Good luck with your analysis!

You May Also Enjoy