Update README.md

README.md

@@ -42,16 +42,18 @@ size_categories:
 English (en)
 
 ## Dataset Structure
-The raw data on this repo contains compressed chunks of newspaper scans for each year. Each scan has it's own JSON file named as the {scan_id}.json. 
-The data loading script takes care of the downloading, extraction and parsing to outputs of two kinds : 
+The raw data on this repo contains compressed chunks of newspaper scans for each year. Each scan has its own JSON file named as the {scan_id}.json. 
+The data loading script takes care of the downloading, extraction, and parsing to outputs of two kinds : 
 
 + Article-Level Output: The unit of the Dataset Dict is an associated article  
 + Scan Level Output: The unit of the Dataset Dict is an entire scan with all the raw unparsed data
 
 ### Data Instances
-Here are some examples of how the output looks like.
+Here are some examples of what the output looks like.
 
 #### Article level 
+
+```
 {
   'article_id': '1_1870-01-01_p1_sn82014899_00211105483_1870010101_0773',
   'newspaper_name': 'The weekly Arizona miner.', 
@@ -60,17 +62,21 @@ Here are some examples of how the output looks like.
   'headline': '',
   'byline': '',
   'article': 'PREyors 10 leaving San Francisco for Wash ington City, our Governor, A. r. K. Saford. called upon Generals Thomas and Ord and nt the carrying out of what (truncated)'
-} 
+}
+```
 
 #### Scan level
+
+```
 {'raw_data_string': '{"lccn": {"title": "The Massachusetts spy, or, Thomas\'s Boston journal.", "geonames_ids": ["4930956"],....other_keys:values}
+```
 
 ### Data Fields
 
 
 #### Article Level
 
-+ "article_id": Unique Id for an assocaited article
++ "article_id": Unique Id for an associated article
 + "newspaper_name": Newspaper Name
 + "edition": Edition number
 + "date": Date of publication
@@ -81,13 +87,44 @@ Here are some examples of how the output looks like.
 
 #### Scan Level
 
-"raw_data_string": Unparsed scan-level data tha contains scan metadata from Library of Congress, all content regions with their bounding boxes, OCR text and legibility classification
+"raw_data_string": Unparsed scan-level data that contains scan metadata from Library of Congress, all content regions with their bounding boxes, OCR text and legibility classification
 
 
 ### Data Splits
 
 There are no train, test or val splits. Since the dataset has a massive number of units (articles or newspaper scans), we have split the data by year. Once the dataset is loaded,
 instead of the usual way of accessing a split as dataset["train"], specific years can be accessed using the syntax dataset["year"] where year can be any year between 1774-1963 as long as there is at least one scan for the year.
+The data loading script provides options to download both a subset of years and all years at a time. 
+
+### Accessing the Data
+
+There are 4 config options that can be used to access the data depending upon the use-case. 
+
+```
+from datasets import load_dataset
+
+#  Download data for the year 1809 at the associated article level (Default)
+dataset = load_dataset("dell-research-harvard/AmericanStories",
+    "subset_years",
+    year_list=["1809", "1810"]
+)
+
+# Download and process data for all years at the article level
+dataset = load_dataset("dell-research-harvard/AmericanStories",
+    "all_years"
+)
+
+# Download and process data for 1809 at the scan level
+dataset = load_dataset("dell-research-harvard/AmericanStories",
+    "subset_years_content_regions",
+    year_list=["1809"]
+)
+
+# Download ad process data for all years at the scan level
+dataset = load_dataset("dell-research-harvard/AmericanStories",
+    "all_years_content_regions")
+
+```
 
 
 ## Dataset Creation
@@ -127,10 +164,10 @@ Not Applicable
 
 ### Social Impact of Dataset
 
- high quality data that could be used for pre-training a large language model to achieve better understanding of historical English and historical world knowledge. 
+ This dataset provides high-quality data that could be used for pre-training a large language model to achieve better understanding of historical English and historical world knowledge. 
  The dataset could also be added to the external database of a retrieval-augmented language model to make historical information - ranging from interpretations of political events to minutiae about the lives of people's ancestors - more widely accessible.
- Furthermore, structured article texts facilitate using transformer-based methods for popular applications like detection of reproduced content, significantly improving accuracy relative to using the existing OCR.  
- Finally, American Stories provides a massive silver quality dataset for innovating multimodal layout analysis models and other multimodal applications. 
+ Furthermore, structured article texts that it provides can facilitate using transformer-based methods for popular applications like detection of reproduced content, significantly improving accuracy relative to using the existing OCR.  
+ It can also be used for innovating multimodal layout analysis models and other multimodal applications. 
  
 ### Discussion of Biases
 
@@ -141,7 +178,7 @@ All content should be viewed as individuals' opinions and not as a purely factua
 ### Other Known Limitations
 
 As a large corpus of news articles, this dataset could hypothetically be used to train a model to generate realistic news articles.
-While this would be acceptable use, outputs from this model would need to be carefully labeled as AI generated to avoid confusion and alert users to the possibility of factual errors.
+While this would be acceptable use, outputs from this model would need to be carefully labeled as AI-generated to avoid confusion and alert users to the possibility of factual errors.
 Additionally, this model should not be finetuned to generate toxic content.
 
 
@@ -153,7 +190,7 @@ Melissa Dell (Harvard), Jacob Carlson (Harvard), Tom Bryan (Harvard) , Emily Sil
 
 ### Licensing Information
 
-The dataset has a CC-BY 4.0 licese
+The dataset has a CC-BY 4.0 license
 
 ### Citation Information